MODULE-

1:I
NTRODUCTIONANDI
MPORTANCE-MEANI
NGOF
STATI
STI
CSANDTHEIRFUNCTI
ON

STATI
STI
CS

Meani
ngofSt
ati
sti
cs

d‘
Thewor St
ati
sti
cs’
hascomef
rom t
heLat
i d‘
nwor st
atus’
,t
heI
tal
i
an
d‘
wor st
ati
sta’
ort
heGer d‘
manwor st
ati
sti
k',
theFr d‘
enchwor st
ati
sti
que’
,eachof
whi
chmeanspol
i
tical
stat
e.

 Inear lyday s,fact sandf i

guresaboutt hef inanci alresources, bi
rthsand
deat hs, ar myst rengt handi ncomewer ecol l
ect edf orthepurposeofef fi
cient
admi nist rationwhi chwascal l
edst at i
st i
csi .
e,any thingper t
ainingt othest ate.
 Nowaday sSt at isticsi snotonl yt hesci enceofst atebutitplaysani mpor tant
roleinal lwal ksofl ifeandi nal l branchofsci ent if
icenqui r
y.Inf act ,stati
stics
hasbecomeoneoft heessent ialtool si nmoder nbiology.
 Usual ly, thewor d' st atisti
cs'car riesdi ffer entmeani ngsdependi ngont he
occasi oni nwhi chi ti sused.
o F ore. g., tmaymeanst
i atisti
caldat awhi chr eferstoquant itati
ve
inf ormat ion, st ati
sti
calmet hodwhi chmeanst hemet hodsdeal ingwi th
quant itativ einf ormat ionorst atistical measur esofasampl e.i.
e. ,
Ar i
thmet icmean, st
andar ddev i
ationet c.ofasampl e.
 Byst atist icaldat a,wemeant heaggr egat eoff act swhi chareaf fect edby
mul tipl
ici tyofcauses, numer icallyexpr essed, est i
mat edtoar easonabl e
standar dofaccur acyandcol l
ect edi nasy stemat icmannerf orapr e-
deter mi nedpur pose.
 Statistical met hodi ncl udescol l
ect i
on, classi f
icat i
on, tabul
ation, present ation,
analy sisandi nt er pret ati
onofdat a.
 Biost atisticsist heappl i
cationofst atistical met hodst othepr obl emsof
biologyi ncludinghumanbi ology ,medi cineandpubl i
chealth.
 Biost atisticsisal socal l
edBi omet rymeani ng" biologicalmeasur ement ".

Funct
ionsofSt
ati
sti
cs

 Presentsfact
sinadef init
eform
 Simpli
fiesmassoff i
gur es
 Facil
it
atescompar i
son
 Helpsinformulatinghy pothesi
s
 Helpsintesti
ngt hehy pothesis
 Helpsinpredict
ion
 Helpsintheformul ati
onofsui tabl
epol
i
cies

Li
mit
ati
onsofSt
ati
sti
cs
  Stati
sti
csdoesnotdeal wit
hthei ndiv
iduals.
 I
tdealsonlywi t
hquantit
ati
vechar act
ers.Howev erquali
tati
vechar
act
erscan
benumer i
call
yex pr
essedandanal yzed.
 Foreg.Int
ell
igenceofstudent
sbymar ksobt ai
ned,povertybyi
ncome
recei
ved.
 Stati
sti
calr
esultsaretr
ueonlyonanav erage.
 Stati
sti
csisonlyoneofthemet hodsofst udy i
ngaproblem.
 Stati
sti
csmaybesomet imesmi sused,ifnotproper
lyint
erpret
ed.

DEFI
NITI
ONS

Popul
ati
on

 Asetorcol
lect
ionofobj
ect
spert
aini
ngtoaphenomenonofst ati
sti
cal
enqui
ryi
srefer
redtoasuni
ver
seorpopulat
ionorcensus.(
e.g.
)animal
sina
far
m.

Sampl
e

 Whenafewuni
tsar
eselect
edfr
om apopulat
ion,
iti
scal
l
edasasampl
e.
(
e.g.
)ani
mal
sofapart
icul
arbr
eedi
naf ar
m.

Var
iabl
e

 Thequanti
tat
iveornumer
ical
char
act
eri
sti
coft
hedat
aiscal
l
edasav
ari
abl
e.
(e.
g.)wei
ghtofananimal
.

Const
ant

 I
tisanumeri
calval
ue,
whichi
ssameforal
ltheuni
tsi
nthepopul
ati
on.(
e.g.
)
no.ofcr
edi
thour
sforB.V.
Scst
udent
s.

At
tri
but
e

 Itr
efer
stot
hequal
i
tat
ivechar
act
eroft
hei
temschosen.(
e.g.
)br
eedofan
ani
mal.

Par
amet
er

 Astat
ist
ical
measureper
tai
ningt
oapopulati
oniscal
l
edasapar
amet
er.
(
e.g.
)mean,standar
ddevi
ati
onofthepopul
ati
on.

St
ati
sti
c

 Astat
ist
ical
measurepert
aini
ngtothesampl
eiscal
l
edasast
ati
sti
c.(
e.g.
)
mean,st
andarddev
iat
ionofthesample.

Cont
inuousv
ari
abl
e
  Ifavar
iabl
etakesanint
ermediat
ev aluebet
weenanyspeci
fi
edi
nter
val
,i
tis
call
edasacontinuousv
ari
able.(
e.g.)thewei
ghtofani
mal.

Di
scr
eteordi
scont
inuousv
ari
abl
e

 I
fav ar
iabletakesonl
yintegral
values,
thenitiscal
l
edasadi
scr
ete(
or)
di
scontinuousvari
abl
e.(e.g.
)no.ofanimalsinafarm.

PRESENTATI
ONOFDATA

I
ntr
oduct
ion

 Classif
icat
ionandt abulat
ionreducethecompl exit
yofvastandcompl icated
stati
sti
caldatabutst il
li
tisnoteasytointerpr
etthetabul
ateddata.Di agrams
andgr aphswi l
lcatchtheeyemor eeasil
yt hantabl
eswhichprovidear rayof
fi
gures.Agl anceov eragraphordiagram will
enableanylayman( wit
hout
stati
sti
calknowl edge)togetanideaaboutt heessenti
alchar
acterist
icsoft he
tabulateddatawi t
houtmuchst rai
noreffort.

FUNCTI
ONSANDLI
MITATI
ONSOFDI
AGRAMSANDGRAPHS

Funct
ions

 I
twi llat
tracttheat tenti
onofal argenumberofpersons.
 Theycar rya“ bi
rds–ey eview”impr essi
oninthehumanmi nd.
 I
tsav esal otofv aluableti
mei fpresentedinaform ofsuitablechar
ts&
graphsinst eadofpagesofnumer i
calfi
gur
es.
 Tof acil
i
tatecompar i
sonbetweent woormor esetsofdat a.
 Predicti
onequat ionscanber epresentedbygraphsandt hesewi l
lbe
much hel pf uli
nf orecasti
ngs.

Li
mit
ati
ons

 Theyareapproximateindi
cators.
 Exactandaccurateinf
ormationscanbeobtainedf
rom or
iginal
tabular
i
nformation.
 Theycannotsubstit
utethetabularinf
ormat
ion.
 Theyfail
t odi
scl
osesmal ldi
fferencewhenlar
gefi
gur
esar einv
olved.

GRAPHI
CALREPRESENTATI
ONOFDATA

 Gr
aphi
calr
epr
esent
ati
oni
sdonewhen t
hedat
aar
ecl
assi
fi
edi
nthef
orm of
 af
requencydistr
ibuti
on.Thedi
ff
erentgr
aphsar
e
o Hi stogr
am
o F requencyPolygon
o F requencyCurve
o Og i
ve
o L orenzCur ve

Hi
stogr
am

 I
tisav er t
icalbardiagram withoutgapbet weent hebars.
 I
tconsi stsofbarser ectedovert hetrueclassint
er v
al,t
heirareas
beingpr oporti
onal t
othef r
equenci esoftherespect i
veclasses.
 Sincethei nter
valsareofequal width,t
heheightofeachbarser v
esasa
measur eoft hecor r
espondingf requency.
 Drawt het wodiagonalsi nthehighestmodal cl
assr ect
anglesatitst
opcorner
tothepr eandpostmodal r
ect anglecornersandt hexco- or
dinat
eofthepoint
ofintersecti
onist hemode.

Fr
equencyPol
ygon

 I
fpointsar
eplott
edwiththexco-ordinateequaltot
hemi dvalueoftheclass
i
nterv
alsandthecor
respondingfr
equenci esastheyco-ordi
nateandthese
poi
ntsarejoi
nedbymeansofast raightli
ne,weobtainfr
equencypolygon.
 Thesepointsar
ethemi dpoi
ntsofthet opofthebarsinthehist
ogram.

Fr
equencyCur
ve

 I
fpoint
sareplott
edwiththexco-
ordi
nateequaltot
hemidvalueoftheclass
i
nter
val
sandt hecor
respondi
ngfr
equenciesastheyco-
ordi
nateandthese
poi
ntsarej
oinedbymeansofasmoot hcurvethenwegetfr
equencycurve.

Ogi
ve

 Thi
siscumulat
ivefrequencycur
ve.
 Thi
scurvei
sobtainedbymaki nguseofcumul
ati
vef
requencyi
nst
eadoft
he
si
mplefr
equency.

Cumul
ati
veFr
equencyDi
str
ibut
ion

 Afrequencydi
stri
butiongivesthenumberofobser vationst hatliei
nanycl ass
i
nterval
whereasthecumul ativefrequencydistr
ibutiongi vest henumberof
fr
equenciest
hatli
ebel owanymar koraboveanygi v enmar k.
 Wh enderi
vedfr
om af r
equencydi str
ibut
ion,t
hecumul at
ivef requency
di
stri
buti
onofoneki ndgivesthenumberofobser vationsl esst hanthelower
boundari
esofthesuccessiv eclassandt hecumul ativefrequencydi stri
buti
on
ofthesecondkindgivesthenumberofobser v
ationst hatexceedt helower
boundari
esoftheclasswhi char erespecti
vel
yknownast hel essthanand
 greaterthancumul ati
vefrequencydi st
ributi
on.
 Ifwedr awf r
equencypolygont ot heabov etwodi st
ributionwegetcumulat
ive
frequencypolygon( l
essthanandgr eaterthan)
.
 Ifwedr awaf requencycurvetot heabov et wodistr
ibutionint
hesamegr aph,
wegetcumul ati
vef r
equencycurv eorOgi ve.
 Th exco- or
dinateofthepointofintersectionoflesst hanandgreat
erthan
cumul ati
vefrequencycurveisthemedi an.

Lor
enzCur
ve

 Thisisamodi ficat
ionoftheOgi vewhent hevari
ablesandthecumulat
ive
frequenciesareexpressedasper cent
ages.
 I
tser vestomeasur etheev ennessofthedistr
ibut
ionandisuseful
inpict
uri
ng
thedistri
buti
onanddi spersionofwealth,
salesandpr of
it
setc.
,

DI
AGRAMMATI
CREPRESENTATI
ONOFDATA

Poi
ntst
obef
oll
owedi
ndr
awi
ngadi
agr
am

 Foreachdi agram, asui tabl eshor theadingshoul dbegiven.

 Itshouldbedr awnt oexhi bitthest atist
ical matterclear
ly.Itshoul dbesuchas
toall
owi tssignif
icantf eat uretobecl earlyshownoutbyadopt ingsuitabl
e
scaleandwi lldependupont hespaceav ail
able.
 Diagram shouldbedr awnaccur atelywitht hehel pofdrawi ngi nstr
ument s.
 Colouringanddi f
fer entmar kingsshoul dbedonewi thpenci lorwi thcolours.
 Diff
erentcoloursormar ksordot tingsar eusedt oshowdi fferentitems.In
suchcasesl egendshoul dbegi venf orthecol umnandi tem i trefers.I
ndoi ng
so,weshoul dseet hatthev isual i
mpr essionconv ey
edbyt hedi agram isnot
i
nanywayaf f
ected.
 Theor i
ginaldataonwhi cht hedi agram hasbeenbasedshoul dbegi ven,if
necessaryfacingt hedi agr am ast hiswi llhelptheobser vertoseet hedetails
withclari
ty.
 Referencetot hesour ceoft het ableshoul dbepr ovi
ded.

Ty
pesofadi
agr
am

 Onedimensional diagram
o L inediagram
o Ba rDi
agram
 Twod i
mensional (or)AreaDi agram
o Pi ediagram
o Sq uar
edi agram andr ectangl
ediagr
ams
 Thr
eedimensi onal (or
)Volumedi agrams
o Cu bes
o Sp her
es,Cy l
indersetc.
 Pi
ctogr
am
 o Act
ual
pict
ures

ONEDI
MENSI
ONALDI
AGRAM

Li
nedi
agr
am consi
sti
ngofcur
vesandl
i
nesaswel
lasbar
s

 Li
nedi agram
o Th i
sr equiresv ertical l
inest obedr awnatequal intervalseachofl ength
propor t
ional tot hemagni tudeoft hev ariableforthedi fferentitems.
o I thasnowi dthandhenceofv erypoorv i
sual eff
ect .
o I tmakescompar isoneasyal t
houghi tislessat t
ract i
ve.
 BarDiagram
o I ti
st hesimpl estofal lstat i
sti
cal diagrams.
o I tconsi stsofbar sofequal widt h(allhor i
zontalorv ertical)standingon
acommonbasel ineatequal intervals,thelengthoft hebar sbeing
propor t
ional tot hemagni tudeoft hev ariablefordifferenti tems.
 Sub-di
v i
dedbardi agr am orcomponentbardi agr am
o So met i
mest hev ariabl
ei scapabl eofbei ngsub- di
v i
dedi nt otwoor
mor ecomponentpar tseachr epr esent i
ngasubv ariable.
o I nthi scase, allthebar sar esubdi v i
dedbyl i
nesi nthesameor derso
thateachsubdi visionr epr esent sthepar tsinmagni tudei nt hesame
scal e.
o Th eyar epr oper lycol ouredormar keddi ffer
entlyforv isual guidance.
o Sma llsquar esshoul dbegi venbel owt hedi agram cont ainingthesame
colourormar ktoshowt heirsi gnif
icance.
 Superi
mposedorMul ti
pl ebardi agram
o Ba rsmaysomet imesbesuper imposedf orcompar ati
v epur pose.
 Percentagebardi agr am
o Wh ent hecomponentpar tsareexpr essedi nper centagesoft hewhol e,
ther esul t
ingbardi agram i scal ledaper cent agebardi agram.
o I nthi scaseal l t
hebar sar eofequal lengt h.

TWODI
MENSI
ONAL(
OR)AREADI
AGRAM

Pi
edi
agr
am

 Circleswi thar eaproport

ionaltothemagnitudesofthedat aar edrawn( i
.e.
)
radiipr oportionalt othesquar
er ootofthemagni t
udeoft hedat aandt he
component s(subv ar
iabl
es)aredrawnwi t
hsect or
spr oporti
onalinar eat o
theirmagni tude.
 Ac irclesubt endsanangl eof360att hecent
0
reandt hisrepresentsthet otal
.
Ther equiredangl eoft hesectorrepr
esenti
ngt hecomponenti scalculated
andar eadi stinguishedbydiff
erentcolour
sormar kingsandkeyf orthis
shoul dbegi ven.
  Itisusual tostartfr
om ahor i
zontalr
adiustotherightandproceedintheanti
-
clockwi sedirecti
ongivingthequantiti
esindescendingorderofmagnitude
exceptt hemi scell
aneouswhi chisshownatt heend.
 Th elengthshav emor ev i
sualeff
ectthanareasandhencei ti
soflessusefor
compar ativepurpose.Itiscommonl yusedt orepr
esentsingl
eobservati
on
withdifferentcomponent s.

Squar
edi
agr
am andRect
angl
edi
agr
ams

 Theirareasshoul dbepr oport

ionaltothemagni t
udesoft hedata.
 Forsquar ediagrams, wewi l
lhavet otakethesquarerootofthegiv enfi
gures
whichwi l
lgivethemeasur ementoft hesidesofthesquare.Byadopt ing
suit
ablescalewecandr awsquar es.
 Inthecaseofr ect angl
ediagrams, ifwetakeequal br
eath(widt
h)f orthe
rectangl
es,thent heareaswi l
lbepr oport
ionalt
ot hel
engthsandhencet he
l
engt hswill
bepr opor t
ionalt
ot hemagni tudeofthegivenvar
iables.

THREEDI
MENSI
ONAL(
OR)VOLUMEDI
AGRAM

 Thesecompriseofcubes, spheres,pr
isms,cyl
inderandbl
ocks.
 Ofthesecubesar emainlyusedandt hei
rsidesaredrawninproport
iont
othe
cuberoot
soft hemagni t
udesoft hedata.
 Theyarepart
icularl
yusedwhent hedatahasav erywider
ange.Insuchacase,
i
twouldbedifficul
ttorepresentthequanti
ti
esev enbysquar
es.

PI
CTOGRAM

 Tabul ardatacanal sober epr

esent edbypi ctogram, cart
ogr am, mapsand
pi
ct uresast hesedev i
cehelpinat tracti
ngt heattent i
ontost ati
sti
calmatter
whi chwhenpr esentedi ntheordinar ydiagrammat icf ormi sv eryoft
enignored.
 Pictogr am ar
edi agramsofpi ctorial orsemi -
pict
orial nat
ur eandar edrawnin
di
f fer
entsizesaccor dingtoscale.Thought heyar eusef ulinat t
racti
ngthe
attentionofthepeopl e,t
heyv er
yof tenl eanont ables, i
gnor i
ngt hepict
ori
al
di
agr ams.
 Th eycannotbemadeuseofwi thcer t
aincompl i
cat eddat a.

COLLECTI
ONOFDATA

Ast
ati
sti
cal
inv
est
igat
ional
way
sbegi
nswi
thcol
l
ect
ionofdat
a.Onecancol
l
ectt
he
dat
aei
therbyhi
msel
forf
rom av
ail
abl
erecor
ds.

 Thedatacol
lect
edbythei
nvest
igatorhimsel
forbyhisagentf
rom t
hesampl
e
orpopul
ati
onarecal
l
edastheprimar ydat
a.
 Thesourcefr
om whi
chonegathersprimarydat
aiscall
edasthepri
mary
 sour
ce.
 Thedatacol
lect
edfr
om t
heavai
labl
esourcesisknownassecondar
ydata.
 Thesourcef
rom whi
chwearegetti
ngsecondarydat
aisknownassecondar
y
sour
ce.

PRI
MARYDATA

Thedat
acol
l
ect
edor
igi
nal
l
yort
hef
ir
sthandi
nfor
mat
ionoff
act
s.

Met
hodsofcol
lect
ingPr
imar
yDat
a

 Directper sonalobser vati

on:Theinvesti
gatorhi mselfgoest othefiel
dofenquiryandcollectsthe
data.
 Indirectper sonalobservati
on:Theinvestigatorcol l
ect
sdat afrom athir
dperson( cal
l
edaswi t
ness)
,
whoknowsaboutt hedatabeinggathered.
 Dat acol l
ectionthroughagent s,
localreportersetc:Herethei nvest
igat
orappointssomeper soncall
ed
agent sorl ocal r
eportersonhisbehalftocol lectinf
ormation.
 Dat acol l
ectionthroughquest i
onnair
es:Thei nvesti
gatorpreparestheneededinformati
onf orthe
particularst udyintheform ofquesti
ons,calledquest i
onnairesandsendst hesamet o
 ther espondent st ocoll
ectdatafrom therespondent s.

Mer
it
sandDemer
it
sofPr
imar
yDat
a

Met
hods Mer
it
s Demer
it
s

Di
rectper
sonal
observat
ion  I
tisveryaccurate. Expensi
vei
nter
msoft
imeand
 I
ntensi
vedetailscan
becoll
ected. money.

Notsui
tabl
ewhent
hef
iel
dof

enqui
ryi
slar
ge.

I
ndi
rectpersonal
observ
ation.  I
tsav
est
ime.  Witnessshouldpossess
t
hor oughknowledgeofthe
f
act sregardi
ngtheprobl
em of
i
nvest i
gati
on.
 Witnessmustbewi ll
i
ngt ogi
ve
i
nformat i
on.

Datacol
l
ect
iont
hrough
agent
sandl
ocalrepor
ter
s  I
tsav
est
ime.  Theagent
swi
l
lcol
l
ect
 et
c.  Lar
geareacanbe i
nformat i
oni
nthei
rown
cov
ered fashi
on.
 On l
yapproxi
materesul
tscanbe
obtained.
 I
tisexpensive.

Datacoll
ect
ionthr
ough
quest
ionnai
res  I
tsav estime.  Itcannotbeusedi ft
he
 I
tislessexpensive. i
nf ormantsareil
li
ter
ate.
 Geogr aphi
call
y  Re sponsemaybepoor .
di
sper sedareacan  Po ssi
bili
tyofvague/
inaccur
ate
becov ered answer s.

Not
e

Ofallt
het y
pesofcol
lecti
onofpri
marydatai
fthequest
ionnai
r esar
efr
amedski
l
lful
l
ysothatitcanbe
answeredeasil
yandifther
eisacompell
i
ngforcethr
oughwhi chwecancol
l
ectthequest
ionnai
rest
hent
he
datacoll
ecti
onthr
oughquesti
onnai
reswil
lbethebestmethod.

SECONDARYDATA

Thedat
acol
l
ect
edf
rom t
heav
ail
abl
esour
cesl
i
kepubl
i
shedr
epor
ts,
document
s,
j
our
nal
set
c.ar
ecal
l
edsecondar
ydat
a.

 Thesourcef
rom whichthesecondarydat
aarecoll
ectedi
scall
ed
assecondar
ysourceofdata.
 Whil
ethepri
marydat aarecol
lect
edforaspeci
fi
cpurpose,t
hesecondar
y
dat
aaregather
edf r
om sourceswhichweredoneforsomeotherpur
pose.

Sour
cesofobt
aini
ngsecondar
ydat
a

 Publ
ishedreport
s/document
sofi
nst
it
uti
ons,
NGOset
c.
 Sci
entif
icj
ournal
s
 Governmentreport
s
 Booksandnewspaper s

Mer
it
sandDemer
it
sofSecondar
yDat
a

Mer
it
s
  I
tsav
est
ime,
labourandmoney
.

Demer
it
s

 Itmaynotbev er
yaccurat
e
 Allt
hedataneededmaynotbeav ai
l
abl
e
 Itmighthav
ebeencol l
ect
edbysomeimpr
opermet
hodsandi
nsome
abnormalcondit
ion.

CLASSI
FICATI
ONOFDATA

Cl
assi
fi
cat
ioni
sthepr
ocessofar
rangi
ngdat
aint
osequencesandgr
oupsaccor
ding
t
othei
rcommonchar
act
eri
sti
csorsepar
ati
ngt
hem i
ntodi
ff
erentbutr
elat
edpar
ts.

Obj
ect
ivesofCl
assi
fi
cat
ions

 Toremoveunnecessar
ydet
ails
 Tobri
ngoutexpli
cit
lyt
hesi
gnifi
cantfeat
uresi
nthedat
a
 Tomakecompar i
sonsanddrawinginfer
ences

Met
hodsofCl
assi
fi
cat
ion

 Numer icalClassif
icat i
on
o Cl assifi
cationofdat aaccor dingt oquant itativ
echar act er s.(e.g)
classi
f i
cati
onofani mal sinaf arm accor dingt otheirwei ght
 Descr i
pti
veCl assifi
cat i
on
o Cl assifi
cationaccor dingt oattribut esi .
e, qualit
ati
v echar acter s.(e.
g).
classi
f i
cati
onofani mal saccor dingt obr eeds
 SpatialorGeogr aphicalCl assi f
icati
on
o Cl assifi
cationaccor dingt ogeogr aphi cal area.(e.
g)di st r
ict -
wi se
li
vestockpopul ationinTami lNadu
 Tempor alorChr onol ogicalCl assif
ication
o Cl assifi
cationaccor dingt otime( e.g)l i
vest ockpopul ationi ndi ff
erent
years
 Classifi
cati
onaccor dingt oclassi nterv alorf requencydi stribut ion
o Wh ent hedat aar egr oupedi ntocl assesofappr opriat ei nt er
v al,
showi ngthenumberi neachcl ass, wegetf requencydi stribution.Thisi
s
call
edgr oupeddat a.
Theor i
ginal datai scal ledrawdat a.
 Thef oll
owingi sthef requencyt ableshowi ngt hedi str
ibutionofchi cksi n
di
fferentweightcl asses.

Wei
ght
(ingm.
)No.ofChi
cks
 36-
40 12

40-
44 25

44-
48 17

48-
52 05

52-
56 06

56-
60 10

Tot
al 75

Ter
msusedi
nFr
equencydi
str
ibut
ion

Cl
assI
nter
valandCl
assLi
mit
s

 Dataarecl assi f
iedorgroupedi ntor
egularinter
val
swiththerangeofvalues
ofthedat a(Cl assInterval
)wi t
hthelowerandupperl i
mitswhichiar
eknown
asClassLi mi t
s.
 TrueClassI nterv al
o Wh ent heCl assInterval
sarecontinuous,i
tiscall
edTrueorIncl
usive
ClassI nterval.
 ApparentCl assI nterval
o Wh ent her eisasmal lgapbetweent heupperboundaryofanyclass
andl owerboundar yofsucessiveclass,t
hentheClassInter
vali
scall
ed
Appar entorExcl usiv
eCl assI
nterval.

Cl
assFr
equencyorFr
equency

 Cl
assf
requencyorf
requencyi
sthenumberofobser
vat
ionsi
nthatcl
ass.

Wi
dth0rLengt
hoft
heCl
assI
nter
val
  Width0rl
engt
hoftheClassInt
erv
ali
sthedi
ff
erencebet
weent
heupper
boundar
yandlowerboundar
yofthesamecl
ass.

Cl
assMar
k

 I
tisthemidpoi
ntoft
hecl
ass.
 I
tisgivenbyhal
foft
hesum oft
hel
owerl
i
mitandupperl
i
mitofanycl
ass.

TABULATI
ONOFDATA

I
tisasy
stemat
icar
rangementofst
ati
sti
cal
dat
aincol
umnsandr
ows.I
tist
henext
pr
ocessofcondensat
ionofdat
aaf
tercl
assi
fi
cat
ion.Tabul
ati
onsi
samechani
cal
par
tofcl
assi
fi
cat
ion.Theobj
ect
soft
abul
ati
onar
e

 Tabl
esaremorecompr ehensiveandint
ell
igi
bleandcarryalast
ingimpressi
on
onthemindofther
eader .
 Tabl
esfaci
li
tat
equickcompar i
sons.
 Tabl
esfaci
li
tat
eeconomyofspace( whil
epresenti
ng)andtime(whiler
eading)
 Rel
ati
onshi
pandotherrelevantchar
acteri
sti
csofitem canbeeasil
ymar ked
outi
ntabul
ateddat
a.

Thef
oll
owi
ngar
ethepoi
ntst
obeconsi
der
edcar
eful
l
yinpr
epar
ingat
abl
e.

 Thet i
tl
eshoul dbeshor tbutcl
earanditshouldgiveaf ul
lideaofi tscont
ents.
 Thecol umnandr owheadi ngsshouldbeselfexplanat ory
.
 Footnotesmaybegi venifabsol
utel
ynecessar y.
 Prominencemaybegi v
ent oimport
antfactsbydifferentmet hodsofmai l
i
ng
andspaci ng.
 Toh avebet t
erclarit
y,spaceshouldbeleftaftereveryf i
vet otenr ows.
 Itt
het abl
ei stakenf r
om secondarydata,iti
sadv i
sabl etogi veasour cenote
fort
het ablement ioni
ngt hesourceforwhicht hedatai scollected.

Ty
pesoft
abl
e

 ReferencetableorGener alTabl e
o Th esetabl econt ai
nagr eatdealofsummar izedi nf
ormat ion.They
appearusual lyattheendoft her eporti
nthef orm ofappendi xes.
 Textorsummar yt ables.
o Th eyareusedt oanalyseorassi stintheanal ysisofclassifi
eddat a.
Theyarei ncludedi nthedi scussi
onoft hebodyoft her eport.
 Stati
sti
caltabl
es
o Th esear especi altablesusedbyst ati
sti
ciansini nt
erpretingtheresult
s
ofstati
sti
cal analysis.Thecommonl yusedt ablesare‘t’tables,ztabl
e,
Ftableetc.,
 RULESI
NFORMI
NGFREQUENCYDI
STRI
BUTI
ON

 Thecl assi nt erv alshoul dbeofequal widthandofsuchsi zet hatthe

char acter i
st i
cf eat uresoft hedi st
ribut i
onar edispl ay ed.
 Classesshoul dnotbet oolarge( or )toosmal l.Iftool arge,itwi l
linvolve
consi der ableer ror sinassumi ngt hatt hemi dpoint soft heclassi ntervalsare
theav erageoft hatcl ass.Ift
oosmal l,therewillbemanycl asseswi thzero
frequency( or )smal l
frequency .Ther ear ehowev ercer tai
nty peofdat a,which
mayr equi ret heuseofunequal orv aryingclassi nter vals.
 Whent herei si r
r egularf l
owofdat aandwi defluct uat i
nggapamongt he
varieties, vary ingcl assintervalsar etobet aken( or )ot herwiset heremaybea
possi bili
tyofcl asseswi thoutanyf requencyorobser vat i
onsf all
ingint hat
categor y.
 Ther angeoft hecl assesshoul dcov ertheent i
rer angeofdat aandt he
classesmustbecont inuous.
 Itisconv eni entt ohav ethemi dpoi ntoft heclassi nt erv alt
obeani nteger.Asa
gener al rule, thenumberofcl assesshoul dbei nt her angeof6- 16andnev er
mor et han30.

FORMATI
ONOFFREQUENCYDI
STRI
BUTI
ON

 MethodofTal
lyMar
ks
 Ar
rayMethod

Met
hodofTal
lyMar
ks

 Fi
rstwehav
etof
ormt
hecl
assi
nter
val
.Thedi
ff
erencebet
weenmaxi
mum and
minimum val
uesi nthecol
lect
eddataar
enotedanditi
stobedivi
dedbyt henumber
ofrequir
edclasses.Thisv
alueshoul
dberoundedof
ftoourconvenience.
 Thenumbe rofrequiredcl
assescanbecal
culat
edusi
ngthefor
mul asuggest
ed
ei
therbySturge’
sr ul
eorYule'
srul
e.
 Sturge’
srul
e

K=1+3.
322l
ogn(
appr
ox.
)

wher
eKi
sthenumberofr
equi
redcl
assesandni
sthenumberofobser
vat
ions.

 Yul
e'
srul
e

K=2.
5xn¼(
appr
ox.
)

wher
eKi
sthenumberofr
equi
redcl
assesandni
sthenumberofobser
vat
ions.

 Af
terf
ormi
ngt
hecl
assi
nter
val
eachshoul
dbewr
it
tenonebel
owt
heot
herandf
or
eachi
tem i
nthecol
l
ect
eddat
aast
rokei
smar
kedagai
nstt
hecl
assi
nter
val
inwhi
ch
 i
tfal
l
s.
 Usual
l
yaf
terev
eryf
oursuchst
rokesi
nthecl
assi
nter
val
,thef
if
thi
tem i
sindi
cat
edby
str
iki
ngtheprev i
ousfourst
rokes,
thus,
makingiteasyt
ocount.
 Thesestrokesarecountedandthi
siscall
edfor
mat i
onoff
requencydi
str
ibut
ionby
themethodoft all
ymarks.

Ar
rayMet
hod

 Anar
rayi
sanor
der
lyar
rangementoft
hedat
abymagni
tudei
ntheascendi
ngor
descendi
ngor
der
.
 Formtheclassint
erv
alasintheprevi
ousmethod.
 Thenarrangethegi
vendataintheascendi
ngorderofmagnit
ude.
 From t
hear r
ay,wewil
lcountthenumberofobservat
ionsbel
ongi
ngt
oeachcl
ass
andt
henwewi
l
lwr
it
e.
 Thi
smet hodisnoteasy,whenthenumberofobser
vati
onsislar
ge.
 Wec anadoptthi
smet hodincases,
wherethenumberofobserv
ati
onsar
elesst
han
50.

MEASURESOFAVERAGE

Needofanav
erage

 Ast
ati
sti
cal
aver
agecondensesaf
requencydi
str
ibut
ionorr
awdat
aandpr
esent
sit
i
nonesi
ngl
erepr
esent
ati
venumber
.
 I
tisasi
ngl
eval
uewhi
chi
sconsi
der
edast
hemostr
epr
esent
ati
veort
ypi
cal
val
uef
or
asetofv
alues.
 Suchav
aluecannei
therbet
hesmal
l
estonenort
hel
argestonebuti
sonewhi
ch
usual
l
yli
essomewher
eneart
hecent
reoft
hegr
oup.
 Thatiswhyanaver
agei
susual
l
yr ef
err
edtoasameasureofcentr
altendency
.
 I
tislocat
edatapoi
ntar
oundwhichmostoftheot
herv
aluestendtoclust
erand
t
her
efor
eiti
sal
sot
ermedasameasur
eofl
ocat
ion.
 I
tisconsi
der
edasameasur
eofdescr
ipt
ionbecausei
tdescr
ibest
hemai
n
char
act
eri
sti
csoft
hedat
a.

Obj
ect
ivesofav
eragi
ng

 Togetasi
ngl
e,summar
yfi
gur
edescr
ibi
ngt
hepr
omi
nentchar
act
eri
sti
csoft
heent
ir
e
gr
oupofdat
a.
 Tof
aci
l
itat
eint
er-
compar
isonofdi
ff
erentphenomena

Di
ff
erentmeasur
esofav
erages

 Thedi
ffer
entaver
agesarebroadl
ycl
assi
fiedint
ot wogroups
 namelymathematical
aver
agesoral
gebraicaveragesandposi
ti
onal
aver
agesor
av
eragesofposi
ti
on.
 Mat
hemat
ical
aver
agesar
ebasedonal
lobser
vat
ionsandar
ecal
cul
atedbyal
gebr
aic
f
ormul
a.Theyar
e
o Arithmet
icMean(
AM)orMean
o Ge ometr
icMean(
GM)
 o Ha rmoni
cMean(HM)
 Posi
ti
onal
aver
agesar
ebasedonf
ewobser
vat
ionsandoccupycer
tai
nposi
ti
on
amongt
heobser
vat
ions.Theyar
e
o Me di
an
o Mode

SUMMATI
ONNOTATI
ONS

 Ifther
eare‘
n’obser
vati
onsandthei
rval
uesaregi
venbyx1,x2,
..
.,
xnthent
he
sum oftheobser
vat
ions,x1+x2+.
..+xncanbewri
tt
enusingthesymbol‘`
(readassi
gma)asfoll
ows:

whi
chmeanst
hesum ofxi'
s,i
taki
ngv
aluesf
rom1t
on

 Whennoambiguit
yisli
kel
ytoar
ise,t
henthesuff
ixicanberemovedandwe
canwri
teasx1+x2+…+xn= x.
 Thenusi
ngthenotat
ionfxi=f
i 1x1+f2x2+…+fnxnwhichcanbewri
tt
en
as fx.

ARI
THMETI
CMEAN

 I
tisthevalueobtai
nedbydi
v i
dingt
hesum oft
heval
uesoft
hegivenitems(of
avari
abl
e)byt henumberofit
ems.Thus,
o Me an=( sum oft
heval
uesofthei
temsint
heseri
es/Tot
al numberof
it
ems)
oIti
susual
lydenotedby .
 Ifwedenoteal
lthe‘
n’observati
onsinaser
iesbyx1,x2,
x3,
………xn,
then
ari
thmeti
cmeanormeanf orthatser
ieswi
l
lbegivenby

 Inthecaseofafr
equencydist
ri
butioni
fthedi
ffer
entcl
assmarksofthe‘
n’
classesar
edenot
edbyx1,x2,…,xnandt
hecorrespondi
ngfr
equenci
esbyf1,
f2,
…, fn,t
henthemeanofthedata i
s
Thus,
A.M =f
xi/
i N wher
exii
sthemi
dval
ueoft
hecl
asswhosef
requencyi
sfiand
N=t
otal
frquency=f
i

Wei
ght
edar
it
hmet
icmean

 I
ncomput i
ngsi mpleAM, i
twasassumedt hatalltheitemsar eofequal
i
mpor tance.Thi smaynotbeal wayst r
ue.
 Wh enitemsv aryi
nimpor t
ancet heymustbeassi gnedwei ghtinproporti
onto
thei
rrelativ
ei mportance.Thus,awei ghtedmeani sthemeanofwei ghted
i
tems.
 I
ncal culati
ngwei ghtedA.M.eachi tem ismul ti
pliedbyi t
swei ghtandthe
productssoder i
vedaresummedup.Thi st otalisdividedbythet ot
alweight
s
(andnotbyt henumberofi tems)t ogett hewei ghtedmean.
 Symbol i
callyifx1,
x2,…,xnarethedifferentitemswi t
hwei ghtsw1, w2,…,wn
respecti
v el
yt hentheweightedmeani sgi v
enby

 I
nf act,AM ofagroupeddataisaweightedaverageoft
heclassmarksi.e.
middlev al
ueoftheclass-
int
erval
whosewei ghtsar
etherespect
ivef
requency
.
 We i
ghtedmeani samuchbet t
ermeasurethanthesimpl
emeani ncasewhen
t
hei temsinaser i
esarenotequall
yimportant
.

GEOMETRI
CMEAN

 Thegeometr
icmeanist
henthrootofpr
oductof‘
n’i
temsofaser
ies.I
fx1,
x2,
……xnar
ethe‘n’
obser
vat
ionsinaseri
esthenGM i
sgiv
enby:

 Tosi
mpl
i
fyt
heabov
ebyt
aki
ngl
ogar
it
hmsonbot
hsi
des,
 Thus,geomet r
icmeani st
heanti
l
ogarit
hm ofthear
it
hmeti
cmeanoft he
l
ogarit
hmicv al
ues.
 Logari
thm ofgeometricmeanistheari
thmeti
cmeanoflogar
ithmi
cvalues.
 I
nt hecaseoff r
equencydist
ri
but
ion(gr
oupeddata)
,GM i
sgivenby

x1f1x2f2….
xnfn)
1/
N
GM =(

whereN=t ot
alf
requency= f
,
ixii
sthemi
dpoi
ntoft
hecl
asswi
th
f
requencyf
.
i

 Si
mpl
i
fyi
ngbyt
aki
ngl
ogar
it
hm onbot
hsi
des,

HARMONI
CMEAN

 Harmonicmeani sthet
otalnumberofit
emsofav ari
abledi
videdbythesum
ofthereci
procal
softheit
ems.Ifx1,
x2,
…x nar
ethe‘n’obser
vationsandHM
repr
esentstheharmoni
cmean, t
hen

 Har
moni
cmeani
sther
eci
procal
ofar
it
hmet
icmeanoft
her
eci
procal
val
ues.
  I
nthecaseofaf
requencydi
str
ibut
ion,
HM i
sobt
ainedbyusi
ngt
hef
ormul
a,

wher
exii
sthemi
dval
ueoft
hecl
asswhosef
requencyi
sf,
iNi
sthet
otal
f
requency
.

MEDI
AN

 I
tisthevaluewhi chhasgotequal numberofobser vati
onsonei t
herside
whentheitemsar ear rangedi ntheascendi ngordescendi ngor derof
magnit
ude.
 Mediandividestheser iesintotwoequal parts,
onepar twillconsistofal
l
var
iabl
eslesst hanmedi anandt heot herpartgreaterthanmedi an.
 Foranungr ouped( orr aw)dat a,
 Median=si zeof[(n+1)/2]t hitem af t
erar r
angingthedat ain
theascendi ngordescendi ngor derofmagni tude,
whenni sodd.
 Median=meanofsi zeofn/ 2t h and[(n/2)+1]t hitem af t
er
ar r
angingt hedat aintheascendi ng ordescendi ngorderof
magni tude, whenni sev en.
 Foragrouppeddat a(orf requencydistri
but i
on),
  wher
eList helowerboundaryoft
hemedianalclass,
Nistot
alf
requency,mis
cumul
ati
vef r
equencyuptothemedianal
class,
ci swidt
hoft
heclassint
erval
andfi
sfrequencyinthemedianalcl
ass.

Not
e

 Inthecaseoff requencydistri
buti
on, medi anist hevaluewhi chhasgotequal
numberoff r
equenciesonei t
herside( i
.e.
)whi chcor respondst othe
cumul ati
vefrequencyofN/ 2.I
tisobt ainedbyf i
ndingtheMedi analclassf r
om
thelessthancumul ati
vefrequencydi str i
but i
on( thelowerboundar yofthe
medi analcl
asscor r
espondst othesi zeoft hecumul at
ivefrequencyjustless
thanN/ 2andupperboundar yofthemedi anal cl
asscor respondingtot hesize
ofcumul ati
vef r
equencyjustgreatert hanN/ 2).
 Me diancanbecomput edusingogi ve.Itist hexcoor di
nateoft hepointof
i
ntersectionofthelessthanandgr eatert hancumul ati
vefrequencycur ve.

MODE

 I
tisthesi zeoft
hemostf r
equenti
tem i
nalargesetofdat
a.Thusmodei st
he
valueofthatvar
iabl
ewhichoccursmostfr
equentlyorr
epeat
sitsel
fthe
greatestnumberofti
mes.
 I
nt hecaseofgroupeddatamodecanbecal cul
atedby

wher
eLi
sthel
owerboundar
yoft
hemodal
class,
f1,
f2ar
ethef
requenci
esi
nthe
pr
ecedi
ngandsucceedi
ngmodal
classandci
sthecl
assi
nter
val
.

Al
ter
nat
ivef
ormul
a

wher
eLi
sthel
owerl
i
mitoft
hemodal
class,
f1,
f2ar
ethef
requenci
esi
nthe
pr
ecedi
ngandsucceedi
ngmodal
classesr
espect
ivel
yandfi
sthef
requencyoft
he
modal
class.
Not
e

 Modecanbecomput edfrom histogram .I

tist
hexcoordi
nateofthepointof
i
ntersecti
onofthetwodiagonalsf r
om thetopcorner
softhemodal cl
assto
thepreandpostmodal cl
asst opcor ners.
 Asaf ir
stappr
oximati
on,mi dpointofthemodal cl
asswil
lbetakenasthe
val
ueoft hemodewhi chiscall
edcr udemode.
 I
namoder at
elyasy
mmet ricaldist
ribut
ion

mean-
mode=3(
mean-
medi
an)
,(appr
oxi
mat
ely
),

Mode=3medi
an–2mean(
appr
oxi
mat
ely
).Thi
sisempi
ri
calmode.

 Adistr
ibutioncanhav emoret hanonemode.I fi
thasgotonemode, i
tis
call
eduni modaldist
ri
buti
on;ifithasgottwomodes, i
tiscal
ledbi-
modal
dist
ri
bution;i
fithasgotthreemodes, iti
scal
ledtri
-modaldi
str
ibut
ion;ifithas
gotmor ethanthreemodes, i
ti scal
ledmult
i-
modal orpol
y-modaldistr
ibuti
on.

PROPERTI
ESOFAVERAGES

Pr
oper
ti
esofar
it
hmet
icmean

 Thesum oft
hedeviat
ionsoft
heit
emsf r
om themeanisequaltozer
o.
 Thesum oft
hesquareddevi
ati
onsfr
om themeanissmall
erthanthesum of
thesquar
eddev
iat
ionsoftheit
emsfrom anyot
herval
ue

 Theproductofthemeanwi
tht
henumberofobser
vat
ionsgi
vest
het
otal
of
theor
igi
naldata,

 I
f ar
ethemeansoft
het
wogr
oupswi
tht
henumberofobser
vat
ions,
n1andn2r
espect
ivel
ythemeanoft
hecombi
nedgr
oup i
sgi
venby

 AM >GM >HM
 Whenal
lthev
aluesar
eequal
,AM =GM =HM
 Forasymmetri
cal
dist
ri
but
ion,
AM =medi
an=mode
  Foraposi
ti
velyskeweddi
str
ibuti
on,AM >medi
an>mode(
shor
t
tai
lont
heleft)
 Foranegat
ivel
yskeweddist
ribut
ionAM <medi
an<mode(
shor
t
tai
lont
heright)

Pr
oper
ti
esofgeomet
ri
cmean

 GM wi
ll
bezer
oifoneormor
eoft
hev
aluesar
ezer
o.
 GM <AM,GM >HM

Pr
oper
ti
esofhar
moni
cmean

 HM <GM <AM

Pr
oper
ti
esofMedi
an

 Meandeviat
iontakenaboutmedianastheor
igi
ni sthemini
mum.
 I
fthedist
ri
buti
onissymmet r
ical
,median=mean=mode,
 Median<mean, andmedian>mode, i
fthedi
str
ibuti
onisposi
ti
velyskewed.
 Median>mean, andmedian<mode, i
fthedi
str
ibuti
onisnegat
ivel
yskewed.

Pr
oper
ti
esofMode

 I
fthedi
stri
buti
oni
ssymmet r
ical
,mode=median=mean.
 I
fthedi
stri
buti
oni
sposi
ti
velyskewed,
mode<medi an<mean,
 I
fthedi
stri
buti
oni
snegat
ivel
yskewed,mode>medi an>mean
 I
fthedi
stri
buti
oni
smoderatelyasy
mmetr
ical
then,
Mode=3median–2
mean(approxi
mat
ely
).

SI
TUATI
ONSWHEREDI
FFERENTAVERAGESAREUSED

 AM i sgenerall
yappl i
cableforal lsor t
sofdat a.Itshoul dbeusedwhent he
distri
buti
oni sreasonabl ysymmet ricalandf ur t
herst atisti
calanal ysi
si stobe
car r
iedoutsuchast hecomput ationoft hest andar ddev iat
ionet c.andal so
al
gebr ai
cmani pulationistobef ollowedsubsequent ly
.
 GM i susedwheni tisdesir
edt ogiv emor ewei ght stosmal li
temsandl ess
wei ghtt
ol argeitemsandi nt hecaseofr at i
os, per centagesand
mi croor
ganisms.
 HM i susedi nav eragingcertaint ypesofr at
iosandr atesandpr oblems
i
nv olvi
ngt i
me.I tgivesmor ewei ghtt osmal l i
tems.
 Themedi anist obeusedwhent heat tr
ibut eoft hedat aar enotdi rectl
y
measur able.Asi tcanbeeasi lylocat edbymer ei nspect i
on,itcanbe
calculat
edwhent hedat aarei ncompl ete.Uset hemedi anwhent he
distri
buti
oni shighl yskewedandt heext remei temsmayhav edistor t
ing
effectsont hemean.
 Modecanbeusedt oknowt hemostt ypical v alueort hemostcommoni tem.
 I
tisal
sousedwhent
hequi
ckestest
imat
eofcent
ral
i
tyi
srequi
red.

MEANI
NGOFDI
SPERSI
ON

 Themeasur
esofcent
ral
tendencyi
ndi
cat
eonl
ythecent
ral
posi
ti
on.Butt
heyhav
e
t
hei
rownl
i
mit
ati
onsanddonott
hrowl
i
ghtont
hef
ormat
ionoft
heser
iesofdat
a.
 Somet
imest
heymayof
fermi
sleadi
ngr
esul
t oo.Foreg.Consi
st dert
hef
oll
owi
ng
t
hreeser
ies.

o Ser
iesA50,
50,
50,
50
o Ser
iesB50,
56,
48,
42,
43,
59,
52
o Ser
iesC1,
29,
120
 Theyhav
ethesamemean50.Hencewemayconcl
udet
hatt
heseser
iesar
eal
i
kei
n
nature.Butacl oseexami nationshallrevealt
hatthedi str
ibutionsdi ff
erwi delyf
rom
oneanot her.
 Inonedi stri
but i
on, t
hev aluesmaybecl osel
ypackedandi nt heot her, t
heymaybe
widelyscat t
ered.Suchav ariationiscaledscat
l t
er,spreadordi spersion.
 Henceanav eragei smor emeani ngfulwhenitisexami nedi nt helightofdi sper
sion.
 Whendi sper sionisnotsi gnificantthent heaverageappear st obeat rue
representati
v ef i
gureoft heser iesandwhendi spersi
oni ssigni fi
cant,iti
mpl i
esthat
theav er
agei sf arfrom bei ngat r
uerepresentati
vef i
gure.
 Themeasur ementoft hescat teri
ngofi t
em inadi str
ibuti
onaboutt heav erageis
call
edameasur eofv ari
ationordi spersi
on.
 Measur esofdi spersi
onal soenabl ecompar i
sonoft woormor edi str
ibutionswi t
h
regardtot heirv ar
iabil
ityorconsi stency

Obj
ect
ivesofmeasur
esofdi
sper
sion

 Todet
ermi
nether
eli
abi
li
tyofanaver
age
 Toser
veasabasi
sforcont
roloft
hev ar
iabi
l
ity
 Tocompar
etwoormoreseri
eswithregardtothei
rvar
iabi
l
ity

Di
ff
err
entmeasur
esofdi
sper
sion

 Range
 Quarti
leDevi
at i
on(Q.
D)
 MeanDev iat
ion(M.D)
 StandardDeviati
on(S.
D)

RANGE

 I
tisthediff
erencebet
weenthehi
ghest(
H)andlowest(
L)valuesintheraw
data.
 Forthegroupeddata,
ther
angeisthedi
ffer
encebet
weent helowerli
mit(
L)of
thefi
rstcl
assandtheupperl
imi
t(H)oft
helastcl
ass.I
tisgivenby

Range=H-
L
 I
tisav er
ysimplemeasureofdisper
sion.
 I
tisusefuli
nthestudyofvar
iat
ioninmoneyrat
eandratesofexchange,
weatherfor
ecastet
c.
 Rel
ativemeasureofdi
spersi
onforrangeist
herat
ioofr
ange(R.R)whichis
gi
venby

R.
R=(
H-L)/(
H+L)

QUARTI
LEDEVI
ATI
ON

 Itisalsoknownassemi -i
nterquar ti
ler
ange.Iti sbasedonquar t
il
eswhi char
e
pointswhi chdi videthedatai ntof ourequalpar t
s.
 Thel owerorf i
rstquar t
il
e(Q1)di videsthelowerhal fofthedi st
ri
buti
onint o
twoequal parts,i.e.
,i
tisthev aluebel owwhi ch25%oft heobser vat
ionli
eand
abov ewhi ch75%oft heobser vationsli
e.
 Similarly,t
heupperort hir
dquar ti
le(Q3)di
v i
dest heupperhal fofthe
distr
ibutioni ntot woequal par ts,i.
e.iti
sthev aluebelowwhi ch75%oft he
obser vati
onsl i
eandabov ewhi ch25%oft heobser vat
ionslie.
 Thedi fference, Q3–Q1i scalledi nterquar
til
er angeand
 QDi sgi venby( Q3–Q1)/2
 Forgr oupeddat a,Q1isthev aluewhi chcor r
espondst ot hecumul at
ive
frequencyofN/ 4andQ3i sthev aluewhichcor respondst othe
cumul ativef requencyof3N/ 4.
 Quar t
ileDev iationi susedint hecaseofopenenddi st
ributi
on.

For
mul
atocomput
eQD

 I
nthecaseofr
awdat
a,af
terar
rangi
ngt
hedat
aint
heascendi
ngor
der

 I
nthecaseoff
requencydi
str
ibut
ionorgr
oupeddat
a
wher
eL1i
sthel
owerboundar
yoft
hef
ir
stquar
ti
lecl
ass,
m1i
sthecumul
ati
ve
f
requencyupt
othef
ir
stquar
ti
lecl
ass,
f1i
sthef
requencyi
nthef
ir
stquar
ti
lecl
ass
andci
sthewi
dthoft
hecl
assi
nter
val

wher
eL3i
sthel
owerboundar
yoft
het
hir
dquar
ti
lecl
ass,
m3i
sthecumul
ati
ve
f
requencyupt
othi
rdquar
ti
lecl
ass,
f3i
sfr
equencyi
nthet
hir
dquar
ti
lecl
assandci
s
t
hewi
dthoft
hecl
assi
nter
val

 Then,

 Relat
ivemeasur
eofQDi
sknownast
hequar
ti
leco-
eff
ici
entofdi
sper
sion
(QC).

MEANDEVI
ATI
ON

 Meandeviat
ionoraveragedeviat
ioni
naseri
esist
heAM ofthedevi
ati
onsof
t
hevari
ousitemsf r
om anaverage(mean,
medianormode)oftheser
ies
t
aki
ngalldeviat
ionsasposit
ive.

wher
eAi
smeanormedi
anormode
 t
h
wher
eAi
smeanormedi
anormodeandxii
smi
dpoi
ntoft
hei cl
asswi
thf
requency
f
i

 Therelati
vemeasureofMeanDeviat
ioni
sknownasmeancoeff
icientof
di
spersionorcoeff
ici
entofmeandevi
ati
onandi
sobt
ainedbydi
vidingtheMD
bytheav er
agefrom whi
chiti
scomputed

STANDARDDEVI
ATI
ON

 I
tisthemostper fectandwidel
yusedmeasur eofdispersi
on.I
tisan
i
mprovedmet hodoverMD.
 I
tistherootmeansquar eofthedeviati
onsmeasuredf r
om themean.
 I
notherwor ds,SDisthepositi
vesquarerootoft
heAM oft hesquar
eoft
he
devi
ati
onofi t
emst akenfr
om AM oft heseri
es.
 I
tisdenotedby` ’(r
eadas'si
gma').

Wher
exii
sthemi
dval
ueoft
hecl
assi
nter
val
whosef
requencyi
sfandNi
i sthet
otal
f
requency
.

 Si
mpl
i
fyi
ngt
heabov
e,
  Shor
tcutmet
hod,

SHEPPARD'
SCORRECTI
ON

 Incomputingthestandarddevi
ati
on,sometimesgroupingerrormayoccuron
accountofgroupingofdataint
odiff
erentcl
asses.Forstat
ist
icaladj
ust
ment
ofthi
sgroupinger r
or,
Sheppardhassuggestedacor r
ecti
onv al
uetobe
deductedfrom t
hev ar
ianceofthegroupeddat
a,whichisgivenby

Coef
fi
cientofVar
iat
ion

 Rel
ati
vemeasureofst
andarddev
iat
ioni
sknownascoef
fi
cientofv
ari
ati
on
(CVorCOV)andisdef
inedasSD/Mean.

 Thus,CVi stheper centagevari

ati
onf r
om t
hemean,wi
thSDbeingtreatedas
thetotal
var i
ation.
 HigherCVi ndicatesgreatervar
iabi
li
tyandl
essCVimpl
iesbet
terconsistency
ofdata.
 VARI
ANCE

 Squareofstandar
ddeviati
oniscal
l
edasvari
ance.Iti
sthemeansquare
devi
ati
on.
 I
tisthesum ofthesquareddevi
ati
onofi
ndiv
idualobser
vat
ionsfr
om t
he
2
meandividedbythenumberofobserv
ati
ons.Iti
sdenotedby.

St
andar
dEr
ror(
SE)

 Themeanofr andom sampl emaybet akenasar epr

esentat
iveofthe
popul
ationmean.
 Thedif
ferencebetweenthesampl emeanandpopul ati
onmeani sdueto
sampli
ngandi ti
scall
edsampl inger
rororstandarderror
.
 Iti
sdefinedastheSDoft hemeanofdi ff
erentsamples,t
akenfrom t
he
popul
ation.
 Ifwestudyonlyonesampl e,t
hen

wher
e‘n’
ist
hesi
zeoft
hesampl
eandSDi
sthatoft
hesampl
e.

Pr
obabl
eEr
ror
(PE)

 PE=2/
3SD(
appr
ox.
)

Pr
oper
ti
esofQD

 QD=(2/
3)SD=PE( appr
ox.
)
 Mean±QDwil
lcov
er50%oft hecases.

Pr
oper
ti
esofMD

 MD=(4/5)SD(
appr
ox.)
 MDaboutmedi
anastheor
igi
nist
hemi
nimum.

PROPERTI
ESOFSD

 SDi sgreat
ert
hanMD, QDandPE
 Meansquar edevi
ati
onwil
lbemi
nimum,ift
hedev
iat
ioni
stakenf
rom AM as
theorigi
n.
 Me an+1SDwi ll
cover68.
27%ofthei
tems
 Me an+2SDwi ll
cover95.
45%ofthei
tems
 Me an+3SDwi ll
cover99.
73%ofthei
tems
  Byaddi
ngorsubt
ract
ingaconst
antf
rom al
ltheobser
vat
ions,
SDi
sunal
ter
ed.

 I
f ar
ethemeansoftwosampl
esofsi
zesn1,
n2r
espect
ivel
ywi
th
SDt
henthecombi
nedSD()i
sgiv
enby

wher
e i
sthecombi
nedmeanand

MERI
TSANDDEMERI
TSOF MEASURESOFDI
SPERSI
ON

Theessent
ial
requi
si
tesofagoodmeasur
eofdi
sper
sionar
ethesameast
hatof
av
erages.

Measur
eofDi
sper
sion Mer
it
s Demer
it
s

Range Iti
seasyt o I
tisnotr
igi
dlydef
ined.I
tisnotbasedonal
lobser
vat
ions.I
tis
calcul
ateand
simpleto af
fect
edmuchbyext
remev
aluesandopenendcl
asses.
understand.
I
tdoesnotpossesssampl
i
ngst
abi
l
ity
.

I
tisnotamenabl
eforf
urt
hermat
hemat
ical
treat
ment
.

Quar
ti
leDev
iat
ion Itiseasyto I
tisnotr
igi
dlydef
ined.I
tisnotbasedonal l
observat
ions.
calcul
ateand
simpleto I
tdoesnotpossesssamplingstabi
li
ty.
understandand
i
snotaf fect
ed Iti
snotamenableforfur
thermathematicalt
reat
ment .
byextreme
i
t emsandopen
endclasses.

MeanDev
iat
ion I
tisbasedonal
lIti
saf
fect
edmuchbyext
remev
aluesandopenendcl
asses.
 obser
vati
ons,it I
tdoesnotpossesssampli
ngstabil
i
ty.
i
srigi
dlydef
ined
andeasyto I
tisnotamenabl
eforfur
thermathemati
cal
treat
ment
.
cal
cul
ateand
si
mpleto
under
stand.

St
andar
dDev
iat
ion Ithasar i
gid I
tisaf
fect
edmuchbyext
remev
aluesandopenendcl
asses.
formulaandi s
basedonal l
obser vations.It
i
scapabl eof
furtheralgebr ai
c
treatment .Iti
s
l
essaf fectedby
sampl ing

MOMENTS

t
h
 Ther momentof‘n’
observ
ati
onsaboutt
hei
rmean ( cal
l
edascent
ral
moments)
,usual
l
ydenotedbyµr(
readasMur)i
sdef
inedas

wher
ex1,
x2,
……xnar
ethe‘
n’obser
vat
ionsand

 I
nthecaseoff
requencyofdi
str
ibut
ion,
 wher
eNi
sthet
otal
frequencyx1,
x2,
…..xnar
ethemi
dval
ueoft
hecl
asses
whosef
requenci
esar
ef1,
f2….
.f
n

Not
e

 I
ncaseofsymmet
ri
cal
dist
ri
but
ion,µ3,µ5,µ7.
..
..
.wi
l
lbezer
o;µr=0f
oral
lodd
v
aluesofr
.

Moment
saboutar
bit
arypoi
ntA
th
 Ther momentof‘
n’obser
vati
onaboutanypoi
ntA(
cal
l
edasr
aw
'
moments)i edbyµ
sdenot risdef
inedas

wher
ex1,
x2….
xnar
e‘n’
obser
vat
ions

wher
ex1,
x,….xnar
ethemi
dval
ueoft
hecl
asseswhosef
requenci
esar
ef1,
f2….f
n

Not
e

1.whenr=0
2.Whenr=1

3.Whenr
=2

'
Rel
ati ween µrand µr
onbet

Usi
ngbi
nomi
nalexpansi
on

SKEWNESS

 I
tisseent hatthemeasuresofcent r
al tendencyindi
catethecent
ralposi
ti
on
orcentr
al t
endencyofthefrequencydi stri
buti
onandt hemeasuresof
di
spersi
ongi veanindi
cati
ont otheext enttowhichtheitemsclusterar
ound
orscatt
erawayf rom t
hecent r
altendency .Butnoneofthesemeasur es
i
ndicat
et heform ort
ypeoft hedistr
ibution.

Consi
dert
hef
oll
owi
ngt
wodi
str
ibut
ions
Cl
ass Fr
equency Cl
ass Fr
equency

0-
5 10 0-
5 10

5-
10 30 5-
10 40

10-
15 60 10-
15 30

15-
20 60 15-
20 90

20-
25 30 20-
25 20

25-
30 10 25-
30 10

Theabov
etwodi
str
ibut
ionshav
ethesamemean(
=15)andSD(
=6)
,buty
ett
heyar
e
noti
dent
ical
dist
ri
but
ion.

 Thedi str
ibuti
onont helefthandsi deissy mmet ri
cal one,whereast he
distri
buti
onont heri
ghthandsi dei sasy mmet ri
cal orskewed.
 Skewnessr eferst othesy mmet ryordepar turefrom sy mmet ry.
 Sy mmet rymeanst hatt henumberofv aluesabov et hemodeandbel owthe
modei ssamei nadat a.
 Whenadi stri
but iondepar tsfrom t hesy mmet ri
cal form, iti
ssaidt obe
asy mmet ri
cal orskewed.
 Evidently,i
nthecaseofsy mmet ricaldistr
ibuti
on, thet wot ail
softhecur veare
ofequal sizeandi nthecaseofasy mmet ri
caldistribution,onetailofthecurve
i
sl ongerthant heother .
 Adi str
ibuti
oni ssai dtobeskewedi nthedi r
ecti
onoft heexcesst ail
.Thusif
ther i
ghttaili
sl ongert hant heleft,thedistri
buti
oni sposi ti
vel
yskewed; ifthe
l
ef ttai
lislongert hant heright,thedi st
ributi
onisnegat i
velyskewed.
 MEASURESOFSKEWNESS

 Pear
soni
anmeasureofskewness
 Bowl
ey’
smeasureofskewness
 Measur
eofskewnessbasedonmoment
s

Pear
soni
anmeasur
eofskewness

 Iti
sgivenby
,mean–mode or3(
mean-medi
an)andt
hecoef
fi
cientof
skewness

Bowl
ey'
smeasur
eofskewness

 Iti
sgivenbyQ3–M –(
M-Q1)
.i.
e.Q3+Q1–2M andt
hecoef
fi
cientof
skewnessis

Measur
eofskewnessbasedonmoment
s

Not
e: eµ
wher 2i
sthev
ari
anceoft
hedat
a
 KURTOSI
S

 I
tisthemeasur
eofthepeakedness.
 I
tindi
cat
esthedegr
eeofflat
nessorpeakednessi
nther
egi
onorar
ear
elat
ingt
othe
modeoff
requencycur
ve.
 Anor
mal cur
veissymmetr
ical
andbel
lshaped.
 Whentwodist
ri
buti
onsar
ecompared,
thetopofthef
requencycur
vemaybef
lat
,
verynarr
oworpeaked.Thischar
acteri
sti
coffr
equencycur
vei
sknownaskur
tosi
sor
peakedness.
 Thenormal cur
veisknownasmesokur ti
c.
 Ac ur
vemor epeakedthannormalcurv
eislept
okur
ti
candthecur
vewhi
chisf
latt
er
thanthenormalcurv
eisplat
ykurt
ic.

Measur
eofKur
tosi
s

β2=µ µ2 wher
eµ
2
 I
tismeasur
edbyβ2-3wher
e, 4/ 4i
sthe4t
hmomentaboutabout
i
sgi
venby
,

 β2=3f
oranormal cur
ve
 β2<3f
orpl
atykur
ti
ccur v
e
 β2>3f
orl
eptokur
ticcurve

Cl
i
ckher
eforGr
aph.
..
 MEASUREOFPROBABI
LITY

 Probabi
l
ityi
sameasur eofchance.
 I
fAi sadesi
redeventand‘r
’isthenumberofoccur
rencesofAoutofthetot
al
numberofpossi
bleoccurr
encessay‘n’
,t
hentheprobabil
i
tyofoccur
renceof
A,denot
edbyP(A),isgi
venby :

P(
A)=(
No.ofoccur
renceofA)/(
Tot
alno.ofoccur
rences)=r/n

 Pr
obabi
li
tyi
sarati
otaki
ngval
uesf rom 0to1.I
tcanneverbenegat
ive.
 I
faneventi
sanimpossi
bleevent,t
hentheprobabil
i
tyi
s0.
 I
faneventi
scer
tai
ntooccur,
t henit
sprobabil
i
tyis1.

Def
ini
ti
ons

 Event
o An yphenomenaoccur ri
ngi nnature
 Equall
yl i
kelyev ent s
o Al lev entshav eequal chanceofoccur rence
 Mutuallyexcl usiveev ents
o I ftheoccur renceofanev entcompl etel
yav oi
dst heoccurr
enceof
anot herev ent,thentheev entsaresai dtobemut ual
lyexcl
usive.
 I
ndependentev ents
o I ftheoccur renceofoneev entinnowayaf fectstheoccurrenceof
anot herev ent,thentheseev entsar esaidtobei ndependent.
 Simpleandcompoundev ent
o Wh ent woormor eev entsoccurt ogether,t
heirhappeningisdescri
bed
asacompoundev ent,whileifonlyoneev enttakesplaceatat ime,i
tis
calledasasi mpleevent.
 Exhaustivecases
o Re f ertoall possibl
eout comeswi thoutanyomi ssions
 Samplespace
o Re pr esentt hesetofal lpossibleout comesofaphenomenon

MATHEMATI
CALANDEMPI
RICALPROBABI
LITY

 Mat hemat i
cal or“apr i
ori
”probabil
ityisone, whichcanbedet ermined
deduct i
velywi thoutanyexperimentationort ri
al.
 Iti
sbasedont heassumpt i
onthatonehasf ullconfi
denceofanev ent
happeni ngoutofsev eralpossi
blealternat
ives( ev
enbefor et
heev ent
happens) ,whi charemut uall
yexclusiveandequal lyl
ikel
y.
 Itassumest hatallcasesar eequall
yl i
kely
,i.e.equal
lyprobableal
l theti
me.
 Howev er,f
ormanyki ndsofchancesandev entsinreall
ifeespeci
allyinsoci
al
andeconomi caspect s, t
hisdefi
nit
ioni sunsuitabl
e,asonecannotdecl area
priorit
hatall casesar eequall
yli
kely.
 Thus, withoutassumi nganyofcer taintyoftheev enthappening,i
fonet ri
esto
 baset hepr obabil
ityofanev entonpastexper ienceofcertai
nout comes
baseduponal ongseriesofexper i
ment s( t
hatis,onthebasisofst at
ist
ical
dat a),thent heprobabili
tyisknownasst ati
sti
calorempi r
icalprobabili
ty.
 If‘n’ist henumberofcasesobser v
edwhi chisnotonlyalargenumberbut
alsoi ncr easesindefi
nitel
yupt oinf
init
ywhi chset sali
mittothepr obabil
it
yof
theev enthappeni ngandwhenanev entAi sfoundt obeoccurringin‘m’
numberofcases, therati
o(m /n)i scloset oP( A),
thest
at i
sti
calprobabil
it
y.
Inot herwor ds,

THEOREM OFTOTALPROBABI
LITYORADDI
TIONTHEOREM

 Theprobabi
li
tyofoccur
renceofoneorotherofasetofmutuall
yexcl
usiv
e
event
sisthesum oftheprobabi
l
iti
esofoccur
renceofthesepar
ateevent
sof
theset
.

Pr
oof

 Let‘n’bet heexhaust i
vemut uall
yexclusi
veway sinwhichal
leventscanoccur.
 Lett hefir
stev entoccurina1oft heseways,thesecondina2oftheseways…
andkthev entoccuri nakoftheseway s.
 Th en,ifp1,p2,
…,pkbet heprobabili
ti
esoftheoccurrenceoft
heseev ent
s,p1=
a1/n; p2=a2/n; p3=a3/n,...
,pk=ak/n.
 Sincet heseev entsaremut uall
yexclusi
ve,t
henumberofway si
nwhi choneor
otherofkev entswilloccuris:

 Supposeifwet aketwomutuall
yexcl
usiv
eevent
sAandB, t
hentheprobabi
li
ty
ofanyoneoft heevent
sAorBwhi charemutual
lyexcl
usiveshal
lhappen,i
s
thesum oftheprobabi
li
tyofoccur
renceofAandtheprobabi
li
tyof
occurr
enceofB

P(
AorB)=P(
A)+P(
B)
 MULTI
PLI
CATI
ONORCOMPOUNDPROBABI
LITYTHEOREM

 Theprobabili
tyofsimultaneousoccur renceofasetofi ndependentev ent
sis
theproductofthesepar at
epr obabil
it
iesoft hoseindependentevents.
 I
fAandBar etwoindependentev entsandt heiri
ndivi
dualprobabil
i
tiesare
P(A)andP( B),t
hent heprobabil
ityofjointoccurr
encesofAandB, i.
et he
probabi
li
tyofthecompoundev entAandB, istheproductofthei
rrespective
probabi
li
ti
es.
 Supposeiftheev entAoccur sinn1way sofwhi chm1way saresuccessfuland
theeventBoccur sinn2way sofwhi chm2way saresuccessful,t
hen

 Thet
otal
outcomeoftheevent
s=n1n2;
 Thet
otal
no.ofsuccessf
ulout
comesf orAandB=m1m2.Ther
efor
e

PROBABI
LITYOFCOMPOUNDEVENTS

 I
fpi
sthepr
obabi
l
ityoft
heev
ent
,thepr
obabi
l
ityt
hati
twi
l
loccuri
nexact
lyxoutofn
snCxqn-xpx.
casesi

Exampl
e

 Anani
mal
Sci
ent
istwi
shest
ochooset
hreeani
mal
sforhi
sresear
ch f
rom 8sheep,
t
hreeofwhichareSuff
olk,
andtheot
herar
eMecher
i.Whati
stheprobabi
l
ityt
hata
r
andomlyselect
edgroupwill
havet
woMecher
iandoneSuff
olksheep?

Thet
otal
possi
blegr
oupof3 =8C3

=8x7x6/
1x2x3

=56

Wecanselectt
woMecher
ioutoff
iv n5C2way
ei sandwecansel
ectoneSuf
fol
k
3
i
n C1ways.

Ther
efor
e,t
hetot
alpossi
blewaysofsel
ect
ing2Mecheriand1Suff
olk=5C2x3C1=
10x3=30.Probabi
li
tyofget
ti
ngtwoMecheriandoneSuffol
kis=30/56=0.54.

 Anani
mal
sci
ent
istwor
kswi
tht
hreedi
ff
erentcat
tl
ebr
eedst
oimpr
ovet
hei
r
ef
fi
ciencyofbeefpr
oduct
ion.Supposehehas10St
eer
s,5ofwhi
char
eAngus,
3ar
e
 Br
ahman,and2areHeref
ordandhesel
ect
sthreeani
mal
satr
andom.Whati
sthe
pr
obabi
l
ityt
hatal
lthet
hreesel
ect
edareAngus?

Thet
otal
numberofway
sofget
ti
ng3ani s=10C3
mal
=10x9x8/1x2x3

=120

No.ofway
sofget
ti
ngt
hreeAngusi
s =5C3=10

Thepr
obabi
l
ityofget
ti
ngt
hreeAngus =10/120=0.
0833

 I
fweassumet
hatmal
eandf
emal
ecal
vesar
eequal
l
yli
kel
y,whati
sthepr
obabi
l
ityof
exact
lyt
hreeoutof10cal
vesbor
nwi
l
lbemal
ecal
ves?

'
p'i
spr
obabi
l
ityofsuccessand'
q'i
sthepr
obabi
l
ityoff
ail
ure.

Then,
probabi
l
ityofget
ti
ng'
x'successi
n'n'
tri
al=ncx.qn-x.px

I
nthegi
vencase,
n=10,
p=q=½;
x=3

Lettheeventofget
ti
ngamal
ecal
fbet
hesuccess,
thenwehav
etof
indt
he
probabi
li
ty,
x=3

10-
3 3
pr
obabi
l
ity =10C3(
½) (
½)

10
=10C3(
½)
=(10x9x8 / 210
1x2x3)x1/

=120/
1024

=0.
1171

PROBABI
LITYDI
STRI
BUTI
ONS-DEFI
NITI
ON

 Theprobabilit
ydistri
butionshowshowt hesetofall possi
blemut uall
y
excl
usiveevent si
sdi str
ibut
ed.
 Theprobabilit
ydistri
butioncanber egardedast hetheoret
icalequivalentofan
empiri
calrelati
vefrequencydistri
buti
on, wit
hitsownmeanandv ariance.
 Ap r
obabil
itydistr
ibutioncomprisesallthevaluesthattherandom v ariabl
e
cantake,withtheirassociat
edpr obabil
i
ties.
 BI
NOMI
ALDI
STRI
BUTI
ON

 Iti
sapr obabi l
itydi stributionexpr essi ngt hepr obabili
tyofonesetof
alt
ernatives,i.e.successorf ailur e.Itisdev elopedundert hef oll
owi ng
assumpt i
ons:
o Ane xper imenti sper for medundert hesamecondi ti
onf oraf ixed
numberoft rials, say’n’
o I neachsucht ri
al ,therear eonl ytwopossi blechancesoft he
exper iment , successorf ailure
o Th epr obabi l
ityofsuccessdenot edby‘ p’whi chremai nsconst antfrom
trialtot r
ial andt hepr obabi li
tyoff ailur
edenot edbyq=1- p
o Th et r
ialsar ei ndependent
 Thisdistributionwasdi scov er edbyJamesBer noulliandhencei tisalso
call
edBer noulliandi stribution.
 Inaser i
esofni ndependentt rai ls,ifpi stheconst antprobabi l
ityofsuccessat
asinglet ri
al,thent hev ariate’x’ , i.
e., t
henumberofsuccessatt hese‘ n’trail
s
i
ssai dtof ollowbi nomi al di
st ribut ion.Thev ariat
et akesv al
uesf r om 0t on( al
l
i
ntegers) ,thepr obabi l
ityofget ti
ng0, 1,2,…, nsuccessesatt hesent r
ail
si s
qn,
nC1qn-1p,nC2qn-2p2,…, nCxqn-xpx, …, pnrespect i
v el
y,whi charet her espective
n
ter
msofbi nomi al expansi on( q+p) .
 Supposei fwehav eNset sof‘ n’ trials,thenumberofset sinwhi chwewi l
l
have0, 1, 2,…, nsuccesswi l
l begi v enbyt hesuccessi vetermsofbi nomi al
n
expansionN( q+p) .Thus, wecl assi fytheset saccor dingtothenumberof
successeswhi cht heycont ainandwegetaf requencydi stri
butionwhi chis
knownast hebi nomi al distri
but i
on.

No.ofsuccess 0 1 2 … x .
..n

Fr
equency NqnNnC1qn-1pNnC2qn2p2 NnCxqn-xpx Npn

Pr
oper
ti
esofBi
nomi
alDi
str
ibut
ion

 Iti
sadi str
ibut
ionofdisconti
nuousordi scr
etevari
ate.
 Ithastwopar ameters(constants).Theyarenandp, wherendenot
esthe
numberofi ndependenttri
alsandpdenot estheconstantpr
obabi
li
tyof
successatasi ngl
etrial
.
 Ittakesval
uesf r
om 0t on(allintegers)
,i.
e.0,1,
2,
…, n
 It'
smeani snpandv ari
anceisnpq,wher eq=1-p,SD= npq);Var
ianceis
alway sl
esst hanmean.
n
 Thedi ff
erentf r
equenci
esarediff
erenttermsofbinomialexpansi
onN( q+p)
 Itissymmet rical,
whenp=q=½
 Whenni slargeandpi ssmallsuchthatnpi sconst
ant,t
hebinomial
distr
ibuti
ontendst oaPoissondistri
buti
on.
 Whenni slargeandp=q=½, thebinomialtendstobecomeanor mal
distr
ibuti
on.
 POI
SSONDI
STRI
BUTI
ON

 Itisadiscret
eprobabili
tydistr
ibut
ionandisl
i
mi t
ingfor
m ofbinomial
distr
ibut
ion,whennisl argeandpi ssmall
suchthatnpisconstant
.
 Itisadistr
ibut
ionofrareev ent
s.Itisal
socal
ledasthelawofimprobabl
e
events.Theequati
oni s:

wher
em i
sthemean,
Nist
het
otal
frequencyandxi
stheno.success

Pr
oper
ti
esofPoi
ssondi
str
ibut
ion

 I
thasonepar ameterm
 I
tisadistr
ibutionofdiscret
evari
ate
 I
ttakesval
uesf rom 0to∞
 Meanisappr oximatelyequalt
ovari
ance
 Skewnessis1/m
 Kurt
osisi
s3+[ 1/m]

NORMALDI
STRI
BUTI
ON

 Binomi alandPoi ssondi stri

but i
onsar ethemor eusef ul theoretical
distr
ibutionsf ordi scretev ariables.Thatis,theyr el
at etot heoccur renceof
disti
nctev ent s.
 Inordert ohav emat hemat icaldistri
buti
onsui t
ablef ordeal ingwi thquant i
ti
es
whosemagni t udei sacont i
nuousv ari
able,acont inuousdi str
ibut i
oni s
needed.
 Thenor mal dist r
ibutionist hemostusef ul t
heoret i
cal distri
butionf or
continuousv ar i
able.
 Itwasdi scov er edbyDeMov i
reast helimiti
ngf or m ofBi nomi al distr
ibution
andwasal soknownt oLapl ace.
 Gaussi st hef irstonewhomader ef
erencet othisandi twaser raticall
ynamed
afterhim asGaussi andi stributi
on.
 Thef requencycur vecor respondi ngtonor mal distri
but i
oni snor mal frequency
curveornor mal curveorGaussi anorLapl acianorpr obabi li
tycur v
e.

Wher
em =mean,
N=t
otal
frequency2=v
ari
ance
Pr
oper
ti
esofNor
malDi
str
ibut
ion

 Itisadi str
ibut i
onofcont i
nuousv ar
iates
 Thev ariatet akesv aluesf rom –∞ t o+∞
 Itisasy mmet rical di stribution, mean=medi an=mode
 Thesl opeoft hecur v ei sbel l shaped.Theendsoft hecurvet
ail
sof
f
asy mpt ot i
callytot hebase
 Ithast wopar amet er sm andσ
 Thef irstandt het hirdquar tilesar eequi-distantfrom t
hemedian
 MD=4/ 5σor0. 7979σ
o me an±2/ 3σcov ers50%oft heobser vati
on
o me an±σcov er s68. 27%oft heobser vat i
on
o me an±2σcov ers95. 45%oft heobser vation
o me an±3σcov ers99. 73%oft heobser vation
 Alloddmoment saboutmean=0
 Skewnessi szer o
 Kur t
osi sis3.I ti smesokur tic.
 Af requencycur v ei sl eptokur ti
cifkur t
osis>3
 Af requencycur v ei spl at y
kur ticifkurtosis<3
 Ifmean=0; SD=1t hent henor mal distri
butioni sastandar
dnormal
distri
but ion

DEFI
NITI
ONOFCORRELATI
ON

 Cor r
elati
onist hest rengthofrelati
onshiportheintensityofassociat
ion
betweent wov ariables.
 Iftwov ari
ablesv aryinsuchawayt hatasonei ncreases(ordecreases),the
otheralsoincreases( ordecreases),thenthecorrelati
onissaidtobeposi ti
ve
ordirect.(
eg.)f eedi ntakeandgr owthrateofanimal s.
 Iftwov ari
ablesv aryinsuchawayt hatasonei ncreases,theotherdecreases
andv iceversa, thent hecorrel
ati
oni ssaidtobeindi r
ectornegative(eg.
)lit
ter
sizeandbi r
thwei ghtofpi gl
ets.
 Ifthereisnor elat i
onshipbetweent hetwov ari
ables,theyaresaidtobe
i
ndependentoruncor rel
ated.

Coef
fi
cientofcor
rel
ati
on

 Ameasur eofcor
relati
onfreefr
om uni
tsofmeasur
ement
siscal
l
ed
coeff
ici
entofcor
relati
on.
 Iti
sdenotedby’r
’,rtakesval
uesfr
om –1to+1.
  Whenr=+1, t
hecor r
elat
ionisper
fectandposi
ti
ve,
r=-
1,t
hecor
rel
ati
oni
s
per
fectandnegat
ive,r=0,ther
eisnocorrel
ati
on

TYPESOFCORRELATI
ON

o Si
mple,par
ti
alandmult
ipl
e
o Li
nearandnon-l
i
near

Si
mpl
e,par
ti
alandmul
ti
ple

 Itisbasedont henumberofv ar i
abl esst udied.Whenonl yt wov ari
ablesar e
i
nv olvedi nthest udyofcor relation, itiscal l
edast hesimplecor rel
ation.
(eg)( i
.)feedi nt akeandgr owt hofani mals,(ii
).bi
rthweightandnumberof
piglets.
 Wh enmor ethant wov ariablesar einv ol
vedi nthestudy,itiseithermul t
ipleor
partialcor r
elation.
 Inmul t
iplecor relation,west udyt hecor relati
onofonedependantv ariable
overal ltheot heri ndependentv ariabl es,(eg.)mi l
kyiel
dv s.fi
rstlactati
on
period, foodsuppl ied,ageet c.
 Inpar ti
alcor relationwest udyt her elat i
onshipbet weentwov ar i
ables,
assumi ngthatt heot hervar i
abl esar econst ant.(eg)corr
elati
onbet weent he
wei ghtofbr oilerandf eedi ntakeassumi ngt heotherfactorslikeareapr ovided,
l
abourused, medi cinalcostet c.asconst ant.

Li
nearandNon-
li
near

 Ont hebasi sofdegr eeofcov ari

ati
on,correlationmaybet ermedasl i
nearor
non-li
near .
 Whent hecor rel
ationbet weent wosetsofv ar i
abl esisper f
ectofdegr eeone
oruni tywhi chmeanst hatthetwov ari
ableshav eanexactf unct i
onal
rel
ationshi p, t
hecor rel at
ionissaidtobel inear .
 Thegr aphical depictionofal i
nearcorrel
ationpr esentsast raightlineandits
functional relati
onshi pi srepresent
edbyt her el ati
on,y=a+bx, wher e‘
a’and
‘
b’ areconst ants.
 Aper fectl i
nearcor r
el ati
onmaybeposi t
iveornegat ive.Thus, it
snumer i
cal
coefficientwi llbeeither+1or–1.Thesear et hel i
mitsofcor r
elati
on.
 Thus, coef fi
cientofcor r
elat
ioncannotbegr eat erthan+1orl esst han–1.If
thecor r el
ationisimper f
ect,i
tsgraphicexposi tionwi l
l benon- l
inear .
 Itwillnotf orm ast raightl i
ne.Non-li
nearcor relationwi ll
alway sbel essthan
unityandi twi l
lli
ebet ween- 1and+1.

METHODSOFSTUDYI
NGCORRELATI
ON

o Scat
terdiagr
am
o Corr
elat
iongraph
o Karl
pearson'
scoeff
ici
entofcor
rel
ati
on
 o Concur
rentdev
iat
ionmet
hod.
o Rankmethod

SCATTERDI
AGRAM

 Ascat terdiagram orscat tergram orscat t

erplotordotdi agram isachar t
prepar edt orepresentgr aphicallyt herelati
onshipbetweent wov ariables.
 Takeonev ariableont hehor izontal andanot heronthev ert
icalaxi
sandmar k
pointscor r
espondi ngtoeachpai roft hegivenobserv ationsaftertaking
suitablescal e.Then,thef i
gur ewhi chcont ainsthecollect i
onofdot sorpoi nt
s
i
scal ledascat terdiagram.
 Thewayi nwhi cht hedotl i
esont hescat t
erdiagram showst hetypeof
correlat i
on.
 Ifthesedot sshowsomet rendei therupwar dordownwar d,
thenthet wo
variablesar ecor rel
ated.Ifthedot sdonotshowanyt r end,ther
ei sabsenceof
correlat i
onbet weent het wov ar
iabl es.

CORRELATI
ONGRAPH

 I
nt hi
smet hod,curv
esar epl
ott
edfort
hedataontwov ar
iabl
es.Byexamining
thedir
ecti
onandcl osenessofthet
wocurvessodrawn,wecani nf
erwhether
ornotthev ar
iabl
esarerelat
ed.
I
fbot
hthecur
vesdrawnont hegr
apharemov
ingin Ont
heotherhand,i
fthecur
vesaremovi
ngintheopposi
te
t
hesamedir
ecti
on(ei
therupwar
dordownward)
, dir
ecti
on,cor
rel
ati
onissai
dtobenegati
ve
corr
elat
ioni
ssaidtobeposi
ti
ve

 Thi
smet hodisnormall
yusedf
orti
meseriesdat
a.However,
li
kescatt
er
di
agram, t
hismethodal
sodoesnotof
feranynumeri
cal
valueforcoef
fi
cient
ofcorr
elat
ion.

KARLPEARSON'
SCOEFFI
CIENTOFCORRELATI
ON

 Oftheseveral
mathemat i
calmethodsofmeasur
ingcorr
elation,t
heKarl
Pearson’
smethod,popular
lyknownasPearsoni
ancoeff
icientofcorr
elat
ioni
s
oft
enused.
 I
tisdenotedbyr.I
tisalsocall
edproductmomentfor
mul a.Itisgi
venby

CONCURRENTDEVI
ATI
ONMETHOD

Thi
smet
hodofst
udy
ingcor
rel
ati
oni
sthesi
mpl
estofal
lthemet
hods.Whati
stobe
f
oundi
nthi
smet
hodi
sthedi
rect
ionofchangeofxandyv
ari
abl
es.

 Thest
epwi
sepr
ocedur
eis:

St
epi
:

o Fi
ndoutt hedir
ectionofchangeofxv ar
iable,i
.eascompar edwit
hthe
fi
rstv
alue,whetherthesecondv al
ueisi
ncreasingordecreasingor
const
ant.Ifi
tisincreasi
ng,puta+sign,i
fitisdecreasi
ng,puta–sign
 andifi
tisconst
ant,putzer
o.Simil
arl
y,ascompar edtosecondvalue,
fi
ndoutwhetherthethi
rdvaluei
sincreasi
ng,decreasi
ngorconstant.
Repeatt
hesamepr ocessfortheotherval
uesalso.Denotethecolumn
asDx.

St
epi
i
:

o I
nt hesameway,f
indoutt
hedi
rect
ionofchangeofyv
ari
abl
eand
denotethi
scol
umnasDy .

St
epi
i
i:

o Multi
plyDxwit
hDyanddeter
minetheval
ueofc,t
henumberof
concurr
entdev
iat
ionsort
henumberofposi
ti
vesi
gnsobt
ainedaf
ter
multi
plyi
ngDxwit
hDy .

St
epi
v:

o Thenappl
ythef
ormul
a

RANKMETHOD

 Someti
meswemaynotknowt heactual
val
ues,
butthei
rrankingmaybe
known.I
nsuchoccasi
ons,
thi
smethodwouldbeofuse.Evenwhentheactual
val
uesareav
ail
abl
e,wecanrankt
hem andmeasur
ecorrelat
ionusi
ngthe
for
mula:

r
s=1-6∑d2)/n(
( n2-1)

wher
e‘d’
ist
hedi
ff
erencei
nther
anki
ngoft
het
woser
iesxandy
,andni
sthe
numberofpai
redobser
vat
ions.

 Whensomeoft hev al
uesar
eequal,givetheaver
agerankandthenadd1/ 12
(m3–m)t o∑d2,wher em st
andsforthenumberofi t
emswhosev aluesare
equal.
 Wh enthenumberofpai r
edobservat
ionsexceed30,iti
sverydi
ff
iculttorank
them andhence,unlessr
ankisgi
ven, i
tisbet
tertoavoi
dthi
smet hod.This
met hodisal
socaledasSpear
l man’srankcorrel
ati
oncoeff
ici
entasi twas
duet othest
ati
sti
cianSpearman.
 PROPERTI
ESOFCORRELATI
ONCOEFFI
CIENT

 Correl
ati
oncoef
fi
cientl
i
esbet ween–1and+1
 Correl
ati
oncoef
fi
cienti
sindependentofchangeofscal
eandor i
ginoft
he
var
iablexandy
 Correl
ati
oncoef
fi
cienti
sthegeomet r
icmeanofthetwor egr
essi
on
coef
fi
ci s†(
ent inmagni
tude)
.

e:† di
Not scussedi
nthenextchapt
er

Coef
fi
cientofdet
ermi
nat
ion
2
 Squareofcorrel
ationcoeffi
cient(r )i
sdefi
nedascoef
fi
cientofdet
erminat
ion.
2
I
fr=0. 85,iti
mpliesthat85%oft hevari
ati
onint
hedependentvari
abl
eisdue
totheindependentvari
ablestudied.

Coef
fi
cientofnon-
det
ermi
nat
ion
2 2
 (
1–r )i
sdef
inedasthecoeffici
entofnon-det
erminat
ion.If1-
r =0.24,we
i
nferthat24%ofthevar
iati
oni nthedependentvar
iabl
ei sduetotheother
v
ariableswhi
charenotstudied.Thisi
scall
edasunexpl ainedvari
ati
ons.

REGRESSI
ON

 Regr essioni st heamountofdependenceofonev ariableont heot her.This

givest her ateofchangeofonev ari
abl ewithrespectt oanot her.
 Themeani ngofr egr essioni stheactofr et
urningorgoi ngback.
 Thist erm wasi nt roducedbyFr ancisGal ton,whenhest udiedt her elati
onship
betweent hehei ght soff atherandsons.
 Wemaybei nterest edi nest i
matingt hev al
ueofonev ariablegiv ent hevalue
ofanot her.Thisi sdonewi t
hthehel pofr egression.
 Thepr oblem ofr egr essioni stofindoutequat ionsoft helines( orcur ves)with
functional relationshi pbet weent wov ari
ables,i
ndependent-xanddependent
-y.
 Thesi mpl el i
nearr egr essionisoft hef orm y=a+bx, wher e‘ b’representsthe
slopeoft hel i
ne( alsocal l
edasr egr essioncoef f
icient)and' a'thei nter
ceptof
thel i
ne.
 Inast udywher edat aonageandwei ghtofani mal sareav ai
lable, agecoul d
beconsi deredast hei ndependentv ari
able,whilewei ghtast hedependent
variable.
 Itmeanst hatwei ghtr egressesonage.
 Inther egr essionl i
ney=a+bx, aandbar ecalculatedby
  biscall
edtheregr
essioncoeff
ici
entoft
hel
ineyonxdenot
edby'byx'
 Theaboveistheregr
essionli
neofyandx.Simi
l
arl
ywehavether
egr essi
on
equat
ionofxonywhi chcanbegi venby
:

x=c+dy
,

 wher edi
stheregr
essi
oncoef
fi
cientoft
her
egr
essi
onl
i
nexony,
denot
edby'
bxy'
whichi
sgiv
enby

PROPERTI
ESOFREGRESSI
ONCOEFFI
CIENT

 Regr essi
oncoeffi
cientgi
vest herateofchangei ntheval
ueofdependent
variable(y)foraunitchangei ntheindependentvari
able(x).
 Regr essi
onequationisusefulinestimati
ngt heval
uesofybyknowi ngthe
valuesofx.
 Bot htheregr
essioncoeffi
cientswi l
lbeofthesamesi gn.
 I
foneoft heregr
essioncoef f
ici
entsisgreaterthanone,theotherwi
ll
bel ess
thanone.
 Pr oductoftheregressi
oncoef fi
cientsi
sthesquar eofcorrel
ati
oncoeff
icient
.

Exer
cise

 Fol
lowingar
ethedat
aonmil
kyi
eld(
inkg/
day
)(y)i
nal
act
ati
onandnumber
ofdays(x)aft
ercal
vi
ng:

Day(
x) 10 14 18 22 26 30 34

Mi
l
kyi
eld(y)1.
7 1.
6 1.
6 1.
5 1.
5 1.
6 1.
5
8 6 2 9 5 0 8
  Fi
taregr
essi
onequat
ionoft
hef
orm y=a+bxandest
imat
ethemi
l
kyi
eldon
t
h
15 dayand23r
dday.

I
NTRODUCTI
ON

I
npr
act
ice,
anyr
equi
redi
nfor
mat
ionmaybeobt
ainedbyf
oll
owi
ngei
thercensus
met
hodorsampl
i
ngmet
hodofdat
acol
l
ect
ion.

 Censusmet hod
o Itreferst othecompl eteenumer ati
onoft hedat afrom eachandev er
y
unitofpopul ationoruni v
erse,i
.e.totalsetofitemsi napar t
icul
ar
investigation.
 Sampli
ngmet hod
o Itisthepr ocessofl earni
ngaboutt hepopul ati
onont hebasisofthe
sampl esdr awnf r
om thepopulation.i.
e.,apartoftheuni v
erseor
popul ati
onsel ectedfrom thepopulationforthepur poseof
investigation.
 Parameter
o Itisthest at i
st icalmeasurepertainingtothepopul ati
on,e.g.,meanof
thepopul ation, SDoft hepopulation,etc,whil
est at
istici
sthestati
sti
cal
measur eper tainingtothesampl e.e.g.,meanoft hesampl e,SDofthe
sampl e, etc.

NEED,
ADVANTAGEANDDI
SADVANTAGEOFSAMPLI
NG

Need

 Inmanysi tuat
ions,censusmet hod,i
nspiteofitsadvant
ages,i
snotpossi
ble
especial
l
ywhenpopul ati
onislargeorinf
ini
te.Ascensusmethodprov
estobe
costl
yand t imeconsumi ng,sampli
ng met hod ofdatacoll
ecti
onappears
appropri
ate.

Adv
ant
age

 Asonl yapar tofthepopul

ationi
sst udi
ed,thismethodsavestime.
 Itsav escostast heamountofl abourandexpensesi nvol
vedwillbel essfor
thesampl e.
 Since onl ya f ew and standard numberofobser vat
ions areinvolved,it
providesr esult
swi t
hgreat
eraccuracy .
 Itprov i
desmor edetai
l
edinformati
onasi tdealswit
honlyfewobservations.
 So met imessampl i
ngistheonl ymet hodav ail
abl
e.Whent hepopul ationis
i
nf i
niteori fart
icl
esaretobedest royedfortest
ingthequali
ty,
censusmet hod
cannotbecar ri
edout.
Di
sadv
ant
ages

 Whatwegetf r
om thatsampleisonlyt
heesti
mateofpopulati
onparameter
whi
chmaydi f
ferf
rom theactualval
ueoft
heparamet
er,yetiti
spossibl
et o
cal
cul
atethesampli
nger r
ororstandar
derr
orwhichi
sgivenal ongwit
hthe
est
imate.

STANDARDERROR/SAMPLI
NGERROR

 I
tisthediff
erencebetweenpar amet
erandstati
sti
c
 I
tisthestandarddeviat
ionofthesamplemeans.Ifonlyonesampl ei
sstudi
ed,
thenst
andar derr
orisSD/ n wheren=sizeofsampl e.
 Esti
mate:isthevalueofpopulati
onpar
ameterobtainedfrom t
hesample.
 Esti
mator:i
st heonewhi chi
susedtoesti
mat ethevalueofthepopul
ati
on
parameter.

Ther
ear
etwot
ypesofest
imat
e:

 Pointesti
mate:isasingleval
uewhichisusedt oesti
matethepopul
ati
on
parameter
 I
ntervalest
imate:i
sani nter
vali
nwhichpopulationparamet
erl
iesbet
ween.I
t
i
sal socall
edfi
ducial
limitorconfi
dencei
nter
v al.

Not
ati
ons

Par
ti
cul
ars Sampl
e Popul
ati
on

Obser
vat
ions y
1,y
2,y
3,.
.. Y1,
Y2,
Y3…

Si
ze n N

Mean

St
andar
ddev
iat
ion s Sor

USESOFSTANDARDERROR
 Standarder rorofthemeani ndi catestheav eragevari
ati
onsinsampl emeans
from thepopul ati
onmean.St andar derrori
susedasani nst
rumentorbasis
fortesti
ngt hehy pothesi
s.
 Th emagni tudeofst andar
der rorgiv esusani deaaboutthereli
abil
i
tyofthe
sampl e.Thegr eaterthestandar derrorisgreatert
hedepar t
ureofactual
v al
ue
from theexpect edv al
ueandhencegr eatertheunreli
abi
li
ty(l
esserthe
reli
abil
it
y)ofsampl e.Thereciprocal ofstandarderrori
stakenasameasur e
ofreli
abili
tyorpr eci
sionofsampl e.Thatis,

 Asni ncreases,SEdecreasesandhencepreci
sionincr
eases.Inl
argesamples,
sampl i
ngdistri
buti
onofstati
sti
csapproxi
matesnor maldi
str
ibut
ion
 Standarderrorhel
psust odetermi
netheli
mi t
s,wit
hinwhichthepopulat
ion
parametersareexpectedt ol
ie.

VESTOCKCENSUS

 Thel i
vestocksect orisimpor t
antt onationaleconomyandconsi deringt he
shortspanofr e-productiv
ityandl if
eofdomest i
canimal ,thefirstCensuswas
organizeddur i
ng1919- 1920andsi ncet heniti sbeingconduct ed
quinquenni ally
(oncei nfi
vey ears)byal ltheSt ates/UTsi nIndia.Sof ar,18
t
h
suchCensuseshav ebeenconduct edandt hel atestonei st he18 Li v
est ock
Censusi nt heser i
eswi th15/ 10/2007ast hedat eofr eference.17thLi vest ock
Censuswasduet obeconduct edi n2002butgotdel ayedandwasconduct ed
withreferencedat e15thOctober ,.Asper17t hl ivestockcensus( 2003) ,
l
ivestockpopul at i
oni nourcount ryis485mi ll
ionsandt hatofpoul tr
yi s489
mill
ions.Indiar anksf i
rsti
nr espectofbuf fal
o, secondi ncat tl
eandgoat s,
thi
rdinsheep, fourthinducks, fi
fthinchi ckensandsi xt
hi ncamel popul ati
on
i
nt hewor ld.Thecount ryhas57%oft hewor ld'sbuf falopopul at i
onThe18t h
l
ivestockcensushasbeenconduct edt hroughoutt hecount r
yandf i
nal drafti
s
th
awaited.18 Li v estockcensusf orI ndiaandTami lNadui sgivenbel ow.

17thLi
vest
ockcensus(
2003)

Li
vest
ock Wor
ld* I
ndi
a

Cat
tl
e 1371.
1 185.
2

I
ndi
genous 160.
5

Cr
ossbr
ed 24.
7
 Buf
fal
o 170.
7 97.
9

Sheep 1024.
0 61.
5

Goat 767.
9 124.
4

Pi
gs 956.
0 13.
5

Horsesand 55.
1 0.
75
Ponies

Donkey
s 0.
65

Camel
s 19.
1 0.
632

Yaks 0.
065

Mi
thun 0.
278

Poul
tr
y 49877 489.
0

Sour
ce:
Dahd.
nic.
in

SAMPLI
NGPROCEDURES

Random sampl
ingorpr
obabi
li
tysampl
ing

 Unrestr
ictedrandom sampli
ngorsimpl
erandom sampl
i
ng
 Restri
ctedrandom sampling
o St rat
ifi
edsampling
o Sy stematicr
andom sampli
ng
o Cl ustersampl
ing
 o Mul
ti
stagesampl
i
ng.

Nonr
andom sampl
ingornonpr
obabi
li
tysampl
i
ng

 Judgmentsampli
ng
 Conveni
entsampl
ing
 Quotasampli
ng

SI
MPLERANDOM SAMPLI
NG(
SRS)

 Simplerandonsample(
SRS)i sthetechni
queofdrawi
ngsamplefr
om the
populat
ion,i
nsuchawayt hateachunitofpopul
ati
onhasanequaland
i
ndependentchanceofbeingincludedint
hesample.I
fNisthepopul
ati
on
sizeandnissamplesize,t
heprobabili
tyt
hatanyoneofNunit
sinpopulat
ion
tobeconsideredasauniti
nthesampl eis1/N.

SRScanbedonebyanyoneoft
hef
oll
owi
ngt
womet
hods

 Lot
terymet
hod
 Methodusi
ngr
andom numbert
abl
e

Lot
ter
yMet
hod

 Thisisav erypopul armet hodofsel ecti

ngrandom samples.
 Inthismethod,al l
thei t
emsofuni versearenumber edornamedonsepar at
e
sli
psofpaper sofi dent i
calshapeandsi ze.
 Thesesl i
psaret henf oldedinthesamemannerandar eputinacontai
ner.
Aftershuffl
i
ng,onesl ipafteranot
heri stakenti
ll
therequi
redsamplesizeis
obtained.
 Thenumberi nthesl ipswoul dconstitut
ethesampleandt hust
heselecti
onof
i
temsdependsent i
r elyonchance.

Met
hodUsi
ngRandom NumberTabl
e

Thelot
terymethodisqui
tecumbersomewhent hesizeoft hepopul
ationi
slarge.
Thealt
ernat
ivemethodofrandom sel
ecti
onisusi
ngt ableofrandom numbers.The
fol
l
owingaresomeoft hetabl
esofrandom numbersav ai
labl
e:

 Tippetsr andom number stableconsi sti

ngof41, 600random uni tsgrouped
i
nt o10,400set sof4di git
sr andom number s
 Fi
sherandYat esr andom numbert ableconsisti
ngof15, 000r andom unit
s
assignedi nto1,500set sof2di git
srandom number s
 Kendall andSmi thrandom number stablehav i
ng10l akhr andom digit
s
groupedi nto25,000set sof4di gitsrandom number s
 RandCor porat
iont ableofr andom number shav ingof1l akhr andom number
digit
sgr oupedinto20, 000set sof5di gi
tedrandom number s
 Rao, MitraandMat thaitableofr andom number sconsi st
ingof20, 000random
 di
git
sgr
oupedi
nto5,
000set
sof5di
git
srandom number
s

Exampl
eofsel
ect
ingsampl
est
hroughr
andom number
s

 Ifthepopulat i
onsi zeis30( N=30)andsampl esizeis15( n=15) ,then,
numbert hepopul ationfrom 1t o30.Since30i satwodigitnumber ,choose
atwodi gitrandom numbert able.
 Inthetwodi gitrandom numbert abl
e,choose01t o90r ej
ecting91-99and00
soast ogi v
eequal chancetoal luni
ts.
 Ha vearandom st artinrandom numbert abl
e.
 Startanywher e, selectthatnumberi fthenumberi sl
essthan30; ifthe
numberi sgr eaterthan30, divideitby30andt aketheremi nder;i
ft he
reminderis0, thatcor respondst o30.Thepr ocesshastober epeatedt i
l
lwe
get15di f
ferentnumber s(i
.e.requir
edsampl esize)
.

ESTI
MATI
ONOFPOPULATI
ONMEANANDCONFI
DENCEI
NTERVALS

 I
fy1,
y …,
2, y
naret
hesampl
eval
ueschosenf
rom t
hepopul
ati
onofsi
zeN,
then
t
hemeanofthesampl
e

 Sampl
emeanwi
l
lbet
akenasest
imat
eofpopul
ati
onmean

Conf
idencei
nter
val

 Forthis,wehavetofi
ndoutstandar
der
rorofpopul
ati
onmean.Est
imat
ionof
standarderr
orofpopulat
ionmean,

 Whenn>30,
95%conf
idencei
nter
val
forpopul
ati
onmean i
sgi
venby
  99%conf
idencei
nter
val
is

 Forsmal
lsampl
es,
ifn<30,
then95%and99%conf
idencei
nter
val
for

STRATI
FIEDRANDOM SAMPLI
NG(
StRS)

 Inthismet hod, thepopul ationofsize‘N’ i

ssubdi v
idedintoadef init
enumber
ofnonov erl
appi nganddi st i
nctsubpopul ationofsi zesN1,N2,…, Nksucht hat
N1+N2+N3+…+Nk=N.Thi sprocedureofdi vi
dingthepopul ationinto
disti
nctsubpopul ati
onsiscal l
edst r
atif
icationandeachsubpopul ati
onis
call
edasast r
atum.
 Wh il
ef or
mi ngast ratum,weseet hattheuni tswi t
hineachst r
at um aremor e
homogenouswi thr espectt othecharacterunderst udy.Betweenst rata,t
here
shouldbegr eaterdi ver
sityorv ar
iabi
li
ty.Thisi sdonewi ththeideaof
i
mpr ov i
ngt hepr ecisionofest i
mat e.
 Wi t
hineachst r
at um ofsize‘ Ni
’,asimpler andom sampl eofniwi ll
bedr awn
sucht hatn1+n2+……+nk=n, whil
e‘n’isthesi zeoft hesampl e.Sucha
sampl i
ngmet hodi scall
edast hest r
atif
iedr andom sampl ing.The' n'sampling
unitschosenar et hestrt
ifiedrandom sampl e(
St .RS)

Si
tuat
ionswher
eSt
.RScoul
dbeused

 Whenthechar
acterunderi
nvesti
gati
onishighlyvari
abl
e
 Whenesti
matesarerequi
redforsubpopul
ationsalso

Not
ati
ons

Par
ti
cul
ars Sampl
e Popul
ati
on
 t
h t
h t
h t
h
Charact
eri
sti
c y
11,
y12… y
i,
jj uni
tini st
rat
um Y11,
Y12… Yij,
j uni
tini st
rat
um
val
ue

Mean y
1,y
2… Y1,Y2…

Adv
ant
agesofSt
rat
if
iedRandom Sampl
ing

 Iti
smor er epresent at
ivethanSRS, asthereisapossi bil
it
yforallsampleunit
s
drawnf rom asi nglecornerofthepopul ati
on.Howev erinSt .
RS,allst
ratums
areequal lyrepr esentedandtherebyav oidthepossi bil
ityofanyessent i
al
groupofpopul ati
onbei ngcompl etel
yeliminatedfrom sampl e.
 St.RSpr ovidesmor eprecisepopul at
ionestimate,besi desestimatesforeach
oft hestratum.
 Asc ompar edwi t
hSRS, theSt.
RSwoul dbeconcent r
at edgeographical
ly,
accor di
ngly ,
timeandmoneyi nvolvedincol l
ecti
ngt hedat aandintervi
ewing
thei ndi
vidualswi l
lbeconsiderablyreducedandsuper v i
sionoffi
eldwor k
coul dbedonewi thgreatereff
icacyandconv eni
ence.

SYSTEMATI
CSAMPLI
NG

 Itisacommonl yempl oyedt echnique,ifthecompl eteanduptodat eli

stof
sampl i
nguni tsareavail
abl e.
 Th i
sconsi stsofsel
ectingonl ythe1stunitatrandom, therestbeingsel
ected
accordingt osomepr edet ermi nedpatterninvol
vingregularspacingofunits.
 Letussupposet hat‘
N’ popul at i
onunitsareseriall
ynumber edfrom 1toNi n
someor derandasampl eofsi ze‘n’
istobechosen, thenletk=N/ nandt hisk
i
scal l
edsampl i
nginterval orsampl i
ngr ati
o.
 Ch ooseanumber( p)atr andom, whichisl esst
hank, thenchooseev ery
kthi
tem, i
.e.,
p,p+k,p+2k, ..
.Thi srandom numberpi scall
edrandom start.

Mer
it
sandDemer
it
sofSy
stemat
icSampl
ing

Mer
it
s

 Iti
smor econveni
entt
hansimplerandom sampl i
ngandstr
ati
fi
edrandom
sampl i
ngasti
meandwor kinvol
vedisrelat
ivel
yless.
 Systemati
csamplingwil
lbemoreeffi
cientthansimpler
andom sampli
ng,
providedt
heli
stfr
om whichsampleunitsaredrawnisknown.

Demer
it
s

 Act
ualsampl
esi
zewil
lbedi
ffer
entf
rom t
hatofrequi
redsi
ze.
 Sampl
emeanisnotanunbi
asedest
imateofpopulat
ionmean.
  I
tmayleadhighl
ybiasedest i
mates,ift
hereareperi
odi
cfeat
uresassociat
ed
wi
thsampli
nginter
val.i
.e.
,ifthel
isthasaperiodi
cfeat
uresandsampling
i
nter
vali
sequaltoormul t
ipleofthatperi
od.

CLUSTERANDMULTI
STAGESAMPLI
NG

Cl
ust
ersampl
e

 Inthiscase,totalpopul ati
oni sdi videddependi ngont heproblem understudy,
i
nt osomer ecognisabl esubdi v i
sionsnamedascl ust
er sandsimpl erandom
sampl eofthesecl ustersi sdr awn.
 Eachandev eryunitinthesel ectedcl ustersisobservedassampl e.
 Forexampl e,Ifwear eint erestedi nobt aini
ngincomeofaci ty
, wholecit
yis
divi
dedintoNdi f
ferentbl ocksorl ocali
tiesandsimpler andom sampl eof‘n’
blocksisdrawn.
 Th eindi
vi
dualsi nselect edblocksf ormt heclustersampl e.

Mul
ti
stagesampl
e

 Insteadofenumer ati
ngal l
sampl euni tsinthesel ectedcl usters,onecan
obt ai
nbet t
erandmor eeffi
cientest i
mat orsbyr esortingt osubsampl ingat
differ
entst ageswi thintheclusters.Thistechni queiscal ledsecondst age
sampl ing,thi
rdst agesampl i
ngandsoon, clustersbei ngt er
medasf i
rststage
sampl inguni t
.
 Theent i
reabov epr ocedureiscalledmul t
istagesampl ing.
 Forexampl e,ifwewantt ost udyconsumpt ionpat t
ernofhousehol dsi nTami l
Nadu.
 Inthef ir
ststage, asimpl erandom sampl eoff ewdi strictswi llbeselected.
 Selecteddi str
ictsthendi vi
dedi nt
ov il
lagesandf rom whi chasi mplerandom
sampl eofv i
l
lagesi sselected,whichiscal ledsecondst agesampl ingunits.
 Inthet hir
dst ageofsampl ing,theselectedv i
l
lageswi llbef urt
herdi vi
dedinto
househol dsandasi mplerandom sampl eofhousehol dswi l
lbesel ectedas
finalsampl eunits.

NON-
RANDOM SAMPLI
NGMETHOD

Pur
posi
veordel
iber
ateorsubj
ect
iveorj
udgmentsampl
ing

 I
tist
heoneinwhi
cht
hei
nvest
igat
ort
akest
hesampl
esexcl
usi
vel
yathi
s
di
scr
eti
on.

Conv
eni
entsampl
ing

 Ift
heinvest
igat
orchoosest
hesampl
esathi
sconv
eni
ence,
iti
scal
l
ed
conveni
entsampli
ng.
Quot
asampl
ing

 Iti
sat ypeofj
udgmentsampl i
ngwherei
nquotasareset
upaccor
dingt
osome
specif
iedchar
acter
ist
icssuchas‘
thi
smuchi nthi
sgroup’
,‘
thi
smuchinot
her
group’andsoon.

I
NTRODUCTI
ON

 Testofsignif
icanceisast
atist
icalprocedur
efoll
owedtotestthesigni
fi
cance
ofthediff
erencebetweenstati
sticsandtheparameterorbet
weenanyt wo
stat
ist
ics.i
.e.betweensamplemeanandpopul at
ionmeanorbet weentwo
samplemeans.

Hy
pot
hesi
s

 Anyst
atementmadeaboutt
hepopul
ati
on

Nul
lhy
pot
hesi
s

 I
tisthehypothesi
sundertestori
nit
ialhy
pothesi
sproposed.
 I
tmayormaynotbet rue.
 I
tisusual
lydenotedbyHo.Nullhy
pothesi
sisneverproved.
 I
tiseit
heracceptedorrej
ectedatsomelevelofsi
gni
f i
cance.

Lev
elsofsi
gni
fi
cance

 Usuallywewi l
lhav etwol
evel
sofsignif
icance.5%and1%l evel
ofsignifi
cance.
 5% levelofsignif
icancemeansthat
,ift
hisexper mentisr
epeated under
i
denticalcondit
ions100times,
thenthechancef orthi
sconcl
usiontogo
wrongi sfi
veoutof100.
 1%l evelofsignif
icancemeansthat
,ift
hisexper mentisr
epeated under
i
denticalcondit
ions100times,
thenthechancef orthi
sconcl
usiontogo
wrongi soneoutof100.

Al
ter
nat
ehy
pot
hesi
s

 Stat
ementcont
rar
ytonul
lhy
pot
hesi
sisal
ter
nat
ehy
pot
hesi
sandi
sdenot
ed
byH1.

Degr
eesoff
reedom(
d.f
)

 Thenumberofobser
vat
ionswhi
char
efr
eet
omov
eorf
reet
ovar
y.

DI
FFERENTTESTSOFSI
GNI
FICANCE

Par
amet
ri
ctest
  Adi
str
ibut
ionwillbeattachedt othetest.Thev ar
iouspar
amet
ri
ctest
sar
e
o No r
mal deviatetestorlargesampl etestorZtest
o Studentsttestorsmal lsampl etest
o Ch i
-squar
et est
o Va r
iance–r ati
ot estorF-test

Non-par
amet
ri
ctest

 I
twi
llbefr
eefr
om dist
ri
buti
onanditi
scalleddi
str
ibut
ionf
reemet
hod.
I
nanytest
,wetakeanyfourt
ypeofdeci
sions:

o Null
hypot
hesi
smaybet ruebutwer ej
ecti
tbyourtestwhichisTypeI
er
ror.
o Null
hypot
hesi
smaybef alsebutweacceptitwhichisTypeIIer
ror
.
o Null
hypot
hesi
sist
rueandweaccepti twhichiscorr
ectdeci
sion.
o Null
hypot
hesi
sisf
alseandwer ejecti
twhichi
scor r
ectdeci
sion.

STATI
STI
CALPROCEDUREFOLLOWEDI
NANYTESTFOR
SI
GNIFI
CANCE

 St
ep1:For
mati
onofnul
lhypothesi
s
 St
ep2:Cal
cul
ati
onoft
eststat
isti
cs

Testst
ati
sti
cs =(St
atist
ics–Par
ameter
)/SEofdif
fer
ence
=(
Diff
erenceint
heval
ueoftwost
ati
sti
cs)/
SEofdi
ff
erence

 Step3:Dependi
ngont
hev
alueoft
estst
ati
sti
cs,
wet
akedeci
sionsdi
ff
erent
fr
om testt
otest.

LARGESAMPLETESTSORZTESTS

 Ztesti
scarri
edout,whensampl
esizei
s>30.Whenn>30,
itwi
l
lfol
l
ow
normaldi
str
ibut
ionwhoseequat
ioni
s

wher
e,m =mean; =st
andar
ddev
iat
ion

Tot
estt
hesi
gni
fi
canceoft
hedi
ff
erencebet
weensampl
emeanandpopul
ati
on
mean

 St
ep1:H0:Nosigni
ficantdi
ff
erencebet
weensampl
emeanandpopul
ati
on
mean
 St
ep2:Testst
atist
icsorZstati
sti
csisgi
venby,

Wher
ex=meanofsampl
e;m =meanofpopul
ati
on

Wher
e =SDofpopul
ati
on;
n=si
zeofsampl
e

When'
s'i
snotknownr
epl
acei
tbywhi
ch'
s'i
sSDofsampl
e.

Then,

 St
ep3:Concl
usion
o |Z|<1.96,wesaythatZisnotsignif
icantandHoisaccepted.We
NS
denot
et hi
sbyZ=() .
*
o |Z|>1.96,Zissi
gnif
icantandHoi srej
ected.Wedenot
et hi
sbyZ=() .
o |Z|>2.58,Zi
shighl
ysignif
icantandHoi srej
ect
ed.Wedenot et
hisas
*
*
Z=() .

Exer
cise

 Asampl eof300broi
lersi
stakenfr
om afar
m andthei
rmeanweightwas
foundtobe1.9kgswi t
hStandar
dDevi
ati
on0.24kgs.Ver
if
ywhetherthe
samplecouldhavebeentakenfr
om apopul
ati
onofmeanweight2.23kgs.

TESTTHESI
GNI
FICANCEOFTHEDI
FFERENCEBETWEENTWO
SAMPLEMEANS
Case-
1:Whent
heyar
etakenf
rom di
ff
erentpopul
ati
ons

 St
ep-
1Ho:Ther
eisnosigni
fi
cantdi
ff
erencebet
weent
wosampl
emeans.
 St
ep-
2:Thet
estSt
ati
sti
csisgi
venby

Wher
e ar
esampl emeanswi
thsi
zesn1,
n2r
espect
ivel
ytakenf
rom popul
ati
on
wi
thstandar
ddeviat
ions1,


 I
fSDoft
hepopul
ati
onsar
enotknown,
repl
acei
tbyt
hesampl
eSD

Wher
e ar
emeanofsampl
eofsi
zesn1,
n2wi
thSDofs1,
s2r
espect
ivel
y

 St
ep-
3:Concl
usi
on-Asi
nthepr
evi
oust
est

Exer
cise

 Themeanmi l
kyiel
dduringalact
ationint
wof ar
msareesti
matedas2176
kgsand2,
425kgswi t
hSD56. 2kgsand43. 2kgsbyt
akingasampl eofsi
zes
50and60respecti
vel
y.Testwhetherthet
wof ar
msdif
fersi
gni
ficantl
yint
hei
r
l
actat
iony
iel
d.

TESTTHESI
GNI
FICANCEOFTHEDI
FFERENCEBETWEENTWOSAMPLE
MEAN

Case-
2:Whent
heyar
etakenf
rom samepopul
ati
on

 St
ep-
1:Ho:Therei
snosi gni
fi
cantdi
ff
erencebet
weent
het
wosampl
emeans
 St
ep-
2:Thetestst
ati
sti
csisgivenby
Wher
e(
SDofPopul
ati
on)i
snotknown,

 St
ep-
3:Concl
usi
on:
asi
nthepr
evi
oust
est

Exer
cise

 Twosamplesofsizes35and48ar et
akenf
rom abr
oil
erfar
m andt
hei
rmean
wei
ght(i
nkgs)wer eworkedoutas1.
8and2.0r
espect
ivel
ywit
hSD0.72and
0.
56.Testt
hesignifi
canceoft
hetwosampl
es.

DI
FFERENCEBETWEENSAMPLEPROPORTI
ONWI
THPOPULATI
ON
PROPORTI
ON

Totestsi
gni
fi
canceoft
hedi
ff
erencebet
weensampl
epr
opor
ti
onwi
thpopul
ati
on
pr
oporti
on

I
t‘n’ist
henumberoft
ri
als,
and‘
p’isthepr
oport
ionofsuccessoutof‘n’t
ri
als,
andP
i
st heexpect
edpr
opor
ti
onofsuccess,t
henpfol
lowsnormal di
str
ibut
ionwhen‘n’
is
l
arge.

 St
ep-
1:Ho:
Ther
eisnosi
gni
fi
cantdi
ff
erencebet
weenpandP.
 St
ep-
2:

 St
ep-
3:Concl
usi
on:
asi
nthepr
evi
oust
est

Exer
cise
  I
naf arm 120calveswer
ebor
ninayearoutofwhich73ar
efemal
e.Testt
he
hypothesi
sthatsexesar
ebor
ninequalpr
oport
ion.

TOTESTTHESI
GNI
FICANCEOFTHEDIFFERENCEBETWEENTWO
PROPORTI
ONS

Tot
estt
hesi
gni
fi
canceoft
hedi
ff
erencebet
weent
wopr
opor
ti
ons

 Ho: Ther
eisnosignifi
cantdif
ferencebet
weent het
wopr oport
ions
 Letp1,p2ar
eobser v
edpr opor
tionsofsuccessoutofn1,
n2trai
ls
respecti
vel
y,
thentheteststat
iti
csis

TOTESTTHESI
GNI
FICANCEOFANOBSERVEDCORRELATI
ON
COEFFI
CIENT

Tot
estt
hesi
gni
fi
canceofanobser
vedcor
rel
ati
oncoef
fi
cient

 Let‘
r’bethesamplecorr
elat
ioncoef
fi
cientand‘’bethepopul
ation
corr
elat
ioncoef
fi
cient
.
 Ho:Thereisnosigni
fi
cantdi
ffer
encebetweencorr
elat
ioncoef
fi
cient
soft
he
sampleandpopulati
on.

 Concl
usi
on:
asi
nthepr
evi
oust
est
.

Not
e

 I
nt hi
scase,wearenoti
nter
est
edaboutrbei
ngsigni
fi
cant
lydi
ff
erentf
rom ,
butweareinter
est
edwhethert
her
eisasigni
fi
cantcor
rel
ati
onbetweenthe
 var
iabl
esstudi
ed.i
.e.whet
herrissi
gni
fi
cant
lydi
ff
erentf
rom 0(
zer
o).
 Hence,i
ntheaboveformul
a,put=o

 Ifzi
snotsignifi
cant, t
henweconcl udethatrisnotsi
gni
fi
canti.
e.t
her
eisno
corr
elat
ionbetweent hev ar i
abl
esst udied.
N.
S
 Wedenot ethisasr=( ) .
 Itzi
ssignif
icant,
thent her eissignifi
cantcorr
elat
ion.
 Then,wedenot ethisasr=( ) .*
 IfZishi
ghlysigni
ficant,thenwesayt hatri
shighlysi
gni
fi
cantandr=( )
**.
 Inbothcases,bothv ari
abl esstudiedarerelat
ed.

Exer
cise

 Thecorr
elat
ioncoef
fi
cientbet
weenfir
stlact
ati
onyiel
dandsecondlact
ati
on
yi
eldof75Tharpar
karcowswasf oundtobe0.5819.Testt
hesigni
fi
canceof
cor
rel
ati
oninthesample.

SMALLSAMPLETESTS

 Whent hesamplesizeislargeandv ari

atesarenormallydist
ri
buted,nor
mal
testisemployedtot estthesignifi
canceofdi f
fer
ences.
 Wi thsmallsamplesandwhent hedegreesoff r
eedom lessthan30, t
he
variatesar
enotnor mallydist
ributed.
 He ncewemakeuseoft he‘t’
di st
ributi
onand‘t’t
estofsignifi
cance.
 Ta bleshavebeenconst r
uctedr el
at i
ngto‘t’
atdiff
erentprobabil
i
tylevel
sfor
variousdegreesoffreedom.

TOTESTTHESI
GNI
FICANCEOFTHEDIFFERENCEOFTHESAMPLE
MEANFROM THEPOPULATIONMEAN

Tot
estt
hesi
gni
fi
canceoft
hedi
ff
erenceoft
hesampl
emeanf
rom t
hepopul
ati
on
mean

 Step1:Ho:Thereisnosi
gni
fi
cantdi
ff
erencebet
weent
hesampl
emeanand
popul
ati
onmean
 Step2:Testst
atist
ici
sgi
venby
 =meanoft
hesampl
eofsi
ze‘
n’wi
thSD‘
s’wi
thd.f
=(n-
1)

m =popul
ati
onmean

 St
ep3:
Concl usi
on
o I
f| t|<tabl
ev alueoftfor(n-1)d.
f.at5%l
ev el tisnon-
signi
ficantand
HO i
saccept ed.Wedenot ethisast=( )N.S.
o I
f| t|>tabl
ev alueoft’f
or(n-1)d.f
.at5%lev el‘t’
issi
gnifi
cant.Hois
rej
ected,wedenot ethi
sast=( ) *.
o I
f| t|>tabl
ev alueof‘
t’
for(n-1)d.f
.at1%l ev eltishi
ghl
ysignifi
cant.
Hoi srej
ected.Wedenot ethisast=( )**

Exer
cise

 Theav er
agefleeceweightofabreedofsheepwasgi venas1700gm per
sheep.From t
hisbreed20sheepwer esel
ectedatrandom andtheaver
age
weightoffl
eecepersheepoft hi
ssamplewasf oundtobe1625gm wi t
hSD
38.75gms.Findwhet hert
hesampl emeanagreeswiththepopulat
ionmean.

TESTOFSI
GNI
FICANCEOFTHEDI
FFERENCEBETWEENTWOSAMPLE
MEANS

 Totestthesignifi
canceoft hedi ff
erencebetweentwosampl emeanswhen
thesi
zesar elessthan30andt heyaredependent(Pai
red‘t’
test)
 Twos amplesar edependentwhent heyhavesomecommonf act
orl
i
nkingt
he
obser
v at
ionsinthetwosampl es.
o Step1: Ho:Thereisnosi gni
ficantdi
ff
erencebetweenthetwosample
means
o Step2: Teststat
ist
icisgivenby

wher
ed=Di
ff
erencei
ntheobser
vat
ionoft
het
wo
sampl
es.s=SDofd
 o Concl
usi
on:
asi
nthepr
evi
oust
est

Exer
cise

 Thefi
rstcl
i
pwool
yiel
dofdaught
ersoft
her
am andt
hei
rdams(
in100gm)
are

Daught
er(x)1 1 2 1 2 1 1 1 820
9 4 1 8 1 7 75

Dam (y
) 9 1 1 1 1 1 98 111
74658 0

 Testt
hesi
gni
fi
cantdi
ff
erencebet
weent
hedaught
erandt
hedam.

NONPAI
REDORUNPAI
RED'
t'
TEST

 Totestthesignifi
canceofthedi fferencebetweent hemeansoft wosampl es,
whent hesampl esarei
ndependent( Non-Pai
redorUn- paired‘
t’test)
 Byindependenceoft hetwosampl es,wemeant hatthereisnor elati
onshi
p
bet
weeni ndi
vi
dualscontri
buti
ngt hesampl e.Thus,thesampl edr awnf rom
di
ff
erentpopul ati
onsordiff
erentpar tsofthesamepopul ati
onwi l
l be
i
ndependent .
o St ep1:Ho: Ther
eisnosi gnifi
cantdiff
erencebet weent hesampl e
means.
o St ep2:Teststati
sti
cs‘t’i
sgi venby
 o Concl
usi
on:
asi
nthepr
evi
oust
est(
wit
hrespectt
od.
f.
)

Exer
cise

 Fol
lowingarethedat
aonf i
rstl
act
ati
ony i
eldofSindhiandSahi
wal.Test
whetherther
eisanysigni
fi
cantdi
ff
erencebetweenthesetwobreeds.

Si
ndhi 66 475464 56665128 5556 2
9

Sahi
wal68 535372 54634133

TOTESTTHESI
GNI
FICANCEOFANOBSERVEDCORRELATI
ON
COEFFI
CIENT

 Tot
estt
hesigni
ficanceofanobservedcor
rel
ationcoef
fi
cient
o Step1:Ho: Ther
eisnosigni
fi
cantcorr
elati
on
o Step2:Teststati
sti
cisgi
venby
  Thecoeffi
cientofcorr
elationbetweenthefi
rstl
act
ati
onyiel
dandsecond
l
actat
ionyieldof25Thar parkarcowsfrom alar
gepopul
ati
onis0.
48.Ist
his
si
gnif
icantofcorrel
ati
oni nthepopulat
ion.

N.
B

 TableA11page557ofst ati
sti
calmethodsbySnedecorandCochr angiv
es
thecrit
icalval
uesofcorrel
ati
oncoeffi
cientt
hatar
esigni
ficantat5%and1%
l
ev el
.If‘n’
istheno.ofobservat
ions,
wehav etoseethetablefordegr
eesof
freedom at(n-2)
.

CHI
-SQUARETESTFORSI
GNI
FICANCE

 Thistesti
susedi nthecasewhenobser vedfrequenci
esar etobetest
edfor
thei
rfitwi
thexpectedortheoret
ical
frequenciesortotestwhethertwo
fact
orsofclassi
ficat
ionofasetofindivi
dualspresent
edi nthefor
m oftwo-
waytableareindependentornot.

Chi
-squar
etestf
orgoodnessoff
it

 Thistestisperfor medt ot estwhet herthedev i

at i
onofobservedf requenci
es
i
nagi vendataf rom theexpect edfrequenciesar eduetorealcausesorduet o
chance.
 Inotherwor ds,thisisat estt odecidewhet hert heobserv
edf requenciesarei
n
accordancewi tht hefrequenci eswi thi
nstatist
icalli
mit
s.Thist estisusedto
decidewhet hert hegivendat ahasagoodf i
twithoneoft heknownf ormsof
distr
ibuti
ons,viz.,Normal ,Binomi alorPoisson.
 Th i
stestisalsousedt ot esttheobser v
ednumberofpr ogeniesi nageneti
c
experimenttof itinMendal i
anl awsofher edity
.
 Th eteststati
sti
cf orgoodnessoff itisgiv
enby
O-obser
vedf
requency
,E-Expect
edf
requency
,df=No.ofcl
asses-
1

I
nthi
scase,

 Ho: Thefi
tisgoodort
her
eisnosi
gni
fi
cantdi
ff
erencebet
weenobser
vedand
expectedf
requency
.

Concl
usi
on

 Ifthecal culatedχ2islesst hant ableχ2 forther espect

ivedegr eesoff reedom
2 2 N.
S
at5%l ev el.χ isnotsi gnifi
cantwhi chisdenot edbyχ =( ) . i
.e.
,Hoi s
accept ed.Thef i
tisgood.
 Ifcalculatedχ2i s>tabl eχ2f orther especti
ved. f.at5%lev elχ2issignifi
cant
edbyχ2=() *
anddenot .Hoi srejected.Thef iti
snotgoodort heoreti
cal
frequenci esar enotaccor dingtot heory .
 Ifcalculatedχ2i s>tabl eχ2f orther especti
vedfat1%l ev elχ2ishighly
2 **
signifi
cantanddenot edbyχ =().Hoi srejected.Thefitisnotgoodor
theoretical f
requenciesar enotaccor dingtot heory.

Exer
cise

 Inafar
m 250cal v
eswerebor nduringaparti
cul
artime.Thenumberofmale
cal
vesis135.Testwhetherthesexesar eequal
lyborn.
 Thefol
lowingtabl
egi
vest heclassi
f i
cat
ionof400plantsaccor
dingt
othe
nat
ureofleavesandfl
owercol our.

Bl
uef
lowerandf
latl
eav
es 23
4

Bl
uef
lowerandcr
impl
edl
eaf 63

Whi
tef
lowerandf
latl
eaf 76

Whi
tef
lowerandcr
impl
edl
eaf27

 Testwhet
hert
hef
requenci
esar
eint
her
ati
o9:
3:3:
1.
 CHI
-SQUARETESTFORI
NDEPENDENCE

 Thisisper formedwhent hedataispr esentedi ntheform ofacont i

ngency
table.
 At ablegivi
ngt hesi multaneousclassificati
onoft hebodyofdat ai ntwo
differ
entway siscal leda“ cont
ingencyt able”Iftherear e‘
r ’
rowsand‘ c’
columns, thet ablei ssaidtobean " rxccont ingencyt able”.χ2testisapplied
totestwhet herthef actorsclassi
fiedar eindependentornot .i.
e.thetwo
factorsareassoci atedornot .
 Inthecont ingencyt ablebothfactorsmaybequal it
ati
v eoronequal i
tati
veand
theot herquant it
ativeorbot hquant i
tative.
 Thedegr eesoff reedom f orrxccont i
ngencyt abl
eis( r-
1)(c- 1).
 Appl i
catonofχ2st
i atist
icsina2x2,χ2cont ingencyt able

 St
ep1:
HO: Thefact
orsar
eindependent
2
 St
ep2:χi sgiv
enby

wi
thdf=(
2-1)(
2-1)=1

 Conclusi
on:asi hecaseofχ2t
nt estf
orgoodnessoffit(wi
tht
herespect
ive
d.f
).
 Theabov ef
ormulaisappli
cabl
ewhenalla,b,c,daregreat
ert
han5.Ifoneor
mor eisl
essthan5,Yat
e‘scorr
ecti
onofconti
nuityi
stobeappli
edwhichisas
fol
lows:

 Inarxcconti
ngencyt
abl
e,t
heexpect
edv
alue(
E)i
nthei
thr
owandj
th
col
umniscal
cul
atedby
wher
e
t
h
o Ri=Sum ofallt
hev al
uesi
nthei r
ow
t
h
o Cj=Sum ofallt
hev al
uesi
nthejcol umn
o N=Gr andtot
ali.
e.,
thesum ofal
l t
heval
uesi
nthegi
vencont
ingency
tabl
e.Then,

 Concl
usi
on:
asi
nthepr
evi
oust
est(
wit
hther
espect
ivedf
).

EXERCI
SE

 1.Thef
oll
owi
ngt
abl
erel
ati
ngt
othesur
viv
alt
oonebr
eedi
ngf
arm dur
ing5
y
ear
s.Testt
heassoci
ati
onoft
hemor
tal
i
tyr
ateandseasons.

Seasonof Sur
viv
edDi
edTot
al
bir
th

Summer 378 82 460

Rai
ny 215 16 231

Wi
nter 204 87 291

Tot
al 797 185 982

 2.Fol
l
owi
ngi
sthev
acci
net
estr
esul
ts.Fi
ndwhet
hert
hei
nocul
ati
onof
v
acci
necanber
egar
dedasef
fect
iveagai
nstt
hedi
sease.

Sur
viv
ed Dead Tot
al
 I
nocul
ated 10 12 22

Non-
inocul
ated 14 13 27

 3.I
nanexper
imentont
hei
mmuni
sat
ionofgoat
sfr
om Ant
hrax,
thef
oll
owi
ng
r
esul
tswer
eobt
ained.Der
ivey
ouri
nfer
enceont
heef
fect
ivenessofv
acci
ne.

Sur
viv
ed Di
ed Tot
al

I
nocul
ated 15 2 17

Noni
nocul
ated 4 10 14

Tot
al 19 12 31

'
F'TESTORVARI
ANCE-
RATI
OTEST

Thi
sist
het
estt
ocompar
ethev
ari
anceoft
wosampl
es

 HO :
Therei
snosigni
fi
cantdi
ff
erencebet
weent
hev
ari
ancesoft
wosampl
es
 The‘F’
Stat
ist
icsi
sgi
venby

S12andS22ar
ethev
ari
anceoft
hesampl
eofsi
zesn1,
n2r
espect
ivel
y.

Fhas2degr
eesoff
reedom,
onef
orhi
gherv
ari
anceandanot
herf
orsmal
l
erv
ari
ance.

N.
B

 Wehav
etoputt
hegr
eat
estofv
ari S12,
ance( S22)
int
henumer
ator

Concl
usi
on

 I
fcal
|F|
<tabFf
ord.
f.=(
n1-
1),
(n2-
1)at5%l
evel
,Fi
snotsi
gni
fi
cant,
denot
ed
 byF=( )N.S.Hoisaccepted.
 Ifcal|F|>t abFf ord.
f.=(n1-
1),(
n2-
1)at5%lev
el,
Fissi
gni
fi
cant
,denot
edbyF
=( )* .Hoi srej
ected
 Ifcal|F|>t abFf ord.
f.=(n1-
1),(
n2-
1)at1%lev
el,
Fishi
ghl
y
signi
fi
cant,denotedbyF=()* *.Hoisrej
ect
ed.

Exer
cise

 Testwhet
hert
hev
ari
ancesar
esi
gni
fi
cant
lydi
ff
erenti
nthef
oll
owi
ngsampl
e.

Sampl
eI(
X1) 45 46 49 25 17 18 13 56 58

Sampl
eII(
X2) 47 49 43 27 29 38 37

DESI
GNOFEXPERI
MENTS

I
ntr
oduct
ion

 Desi gninganexper imentmeanspl anninganexper i

mentsot hatinformati
on
collectedwillberelevanttothepr oblem underinvesti
gation.
 Th eDesi gnofanexper imentisthecompl etesequenceofst epstakenbefore
exper imenti
ngtoensur ethattheappr opri
atedatawi l
lbeobtainedinaway
thatf urni
shanobj ecti
veanalysisleadingtov al
idinferenceswithrespectt
o
thest ateandpr oblem.
 Th epur poseofanyexper imentaldesi gnistoprovidemaxi mum informati
on
rel
ev anttothepr oblem underinvestigati
on

DEFI
NITI
ON

Tr
eat
ment

 Whatweappl
yont hesubjectofinv
esti
gati
oniscall
edtr
eat
ment.e.
g.
Appl
i
cat
ionoffeedtoanimals,appl
icat
ionoff
ert
il
izert
oagri
cul
tur
alpl
ot,
etc.

Exper
iment
almat
eri
alandexper
iment
aluni
t

 Theindiv
idual
orgroupofindi
vi
dualst
hatwillbesubj
ect
edtoatr
eat
menti
s
cal
ledtheexperi
mentaluni
tandthecol
lect
ionofsuchunit
swi
ll
be
experi
mentalmater
ial
.

Response

 Out
comeofanexper
iment
.i.
e.t
het
reat
mentef
fectav
ail
abl
efr
om t
he
 exper
iment
aluni
ts.

Exper
iment
aler
ror

 I
tist
heunit-
to-
unitvar
iat
ionwithi
nthesamet reatmentgroup.Thi
sisa
measur
eofv ar
iati
onduet ouncontr
ollablecauses.I
tdescri
besthefai
l
ureof
t
woidenti
cal
lytreat
edexperimentalunitstoyi
eldident
icalr
esult
s.

BASI
CPRI
NCI
PLESOFEXPERI
MENTALDESI
GNS

 Randomizati
on
 Repli
cat
ion
 LocalContr
ol

Randomi
zat
ion

 I
tisadev
icef
orel
i
minat
ingbi
as.i
.e.byr
andoml
yassi
gni
ngt
reat
ment
stot
he
experi
mentaluni
tsweav oi
dpersonal
bias.Iti
nvolvesgiv
ingequalchancest
oallt
he
experi
mentaluni
tstobesubject
edtodiff
erenttr
eat ment
s.Itcanbedonebytheuse
ofrandom numbertabl
eorbydrawinglots.Therandomizati
onprocedurei
sdif
fer
ent
fordi
ffer
entdesi
gns.

Pur
posesser
vedbyr
andomi
zat
ion

 I
tav
oidspersonal
bias
 I
tmakesthetestval
id

Repl
icat
ion

 I
tist
her
epet
it
ionoft
hesamet
hing,
thati
sbyr
epl
i
cat
ionwemeant
henumberof
experimental
unit
srecei
vi
ngapar t
iculartr
eatment
.Ifanexperi
menthaveequal
repl
icati
onforal
lthetr
eatmentsst
udi ed,t
henthedesigniscal
l
ed“Equi
-r
epli
cat
ed
desi
gn”I fadesi
gnhasgotanunequal r
epli
cat
ionfordif
fer
enttr
eat
mentstheni
tis
knownasadesi gnwithunequalr
eplicati
on.

Pur
posesser
vedbyr
epl
i
cat
ion

 I
tprovi
desanest
imateofexper
imental
err
or
 I
tenabl
esustoobtai
namor epreci
seesti
mat
eoft
hemeanef
fectofanyf
act
or,
si
ncepr
eci
sioni
s1/ S.
E.andS.
E.=S.D./n.As‘n’
increasespr
eci
sioni
ncr
eases.
Themoret
herepl
icati
on,t
hemorethepreci
sionwi
llbe.

Localcont
rol

 I
tref
erst
othebal
anci
ng,
blocki
ngandgr
oupi
ngoft
heexper
iment
aluni
tst
hati
s
employedi
ntheexperiment.I
tref
erstot heski
ll
full
ogicalwayofgroupi
ngthe
experi
ment
aluni
tsinsuchamannert hatther
eismor eunifor
mitywit
hinthesame
groupandt
herei
sgr eaterv
ariabi
l
itybetweendiffer
entgroups.
Pur
posesser
vedbyl
ocal
cont
rol

 Tomaket
heexperi
mental
desi
gnmoreeff
ici
ent
 Tomaket
hetestpr
ocedur
emorepower
ful

CRI
TERI
AFORMAKI
NGBLOCKS

 Char actersthathav einf l

uenceont heresponsev al
uebef orethe
commencementoft heexper imentareconsider edast hecrit
eriaformaki ng
blocks.
 Fore. g.inmi l
ky i
eldst udies,thestageoflactationcanbeconsi deredas
criteri
aformaki ngbl ocksandi nweightgainst udy,wecanconsi dertheinit
ial
wei ghtoriniti
alageorbr eedorsexascr i
teri
af ormaki ngblocks.
 Incompl et
elyrandomi seddesi gn(CRD),thereisnol ocalcontrolapplied,
whi leinrandomi sedbl ockdesi gn(RBD)local contr
ol i
sappl i
edi none
directionwithonecr iter i
onandi nLati
nsquar edesign( LSD)int wodi r
ections
witht wocr i
teri
a.
 Be sidestheset hreepr incipl
esofexperiment aldesi
gns, ther
ear eauxiliar
y
variableandcont r
olthatshoul dbeconsideredwhi l
eexper iment i
ng.

OTHERPRI
NCI
PLESOFEXPERI
MENTALDESI
GNS

Auxi
li
aryVar
iabl
e

 Somechar acter
sarenotbeingalter
edbythet r
eatmentsappli
edbutmay
haveinfl
uenceonthecharacter
sunderstudy .
 Suchcharactercanbeuti
li
sedtoi mprov
et hepreci
sionoft
heestimat
eand
theeffi
ci
encyofdesigns.
 Theyaresochosent hatt
hecollecti
onofinfor
mat i
onaboutthi
sdonotinv
olv
e
costandlabour.

Cont
rol

 Whennot reatmentisappliedov eragroupofexperimentaluni

tsweconsi
der
theseunit
st oconsti
tut
eacont r
ol gr
oup.
 Th epur
poseoft hi
scontrol i
stomakeef f
ecti
vecompar i
son.
 Wh enev
eranexper imentisconduct edtomaker ecommendat i
onofanew
treat
ment,iti
sbettertoi
ncl udeacont r
olgroupasoneoft hetreat
mentgr
oup.

COMPLETELYRANDOMI
ZEDDESI
GN(
CRD)

 Thi
sissimpl
estofal
lexper
imental
desi
gns.
 Thi
sisthedesi
gninwhicht
hetreat
mentsar
eassi
gnedcompl
etel
yatr
andom
 totheexper i
ment al uni
tsorv iceversa.i.e.itimposesnor estr
ict
ionsonthe
all
ocationoftreatment stot heexper i
ment alunit
s.
 CRDi spreferr
edwhenal lt
heexper iment alunit
sconsideredforthe
experimentareknownt obehomogeneous.
 Anynumberofexper i
mental unit
sandt reatment scanbeut i
l
izedinthi
s
design.
 Ast hedesignishi ghlyfl
exibleandsi mpl e, t
heCRDi swidel
yused.
 Analysisissi
mpl e, evenifcertainvaluesar emissing.
 Theexper imentaluni t
swi l
lbeal l
ottedt ot hediff
erenttr
eatmentsbyusing
random numbert abl eorbyl otterymet hod.

COLLECTI
ONANDANALYSI
SOFDATA

 Afterhavi
ngrandomi sedtheexperimentalunit
sov erdi
fferenttreatment s,
i
nitialr
ecordi
ngs(ifany)oftheexperimentalunit
sarenot edagai nsteach
experimentaluni
t.
 Forexampl e,
incaseofwei ghtgainstudy,i
niti
alweightsaret ober ecorded.
 Thent heexperi
ment aluni
tsaresubject
edt orespecti
vetreatment sandaf ter
theexperimentalperi
od,theresponsevalueswillbeobser ved.Thedat awi l
l
bet abul
atedasfoll
ows.

Dat
aofr
esponsev
alues

Tr1 Tr2 .
.. Trt

y
11 y
12 .
.. y
1t

y
2t y
2t .
.. y
2t

. . .
.. .

. . .
.. .

. . .
.. .

. . .
.. .

y
n1t y
n2t .
.. y
nt
t

T1 T2 .
.. Tt
  Letushav
e’t‘
tr
eat
ment
seachhav
ingr
epl
i
cat
ionsn1,
n2…nt,
then,
n1+n2+
..
.+nt=N

STEPWI
SEPROCEDURE

Ho:
Ther
eisnosi
gni
fi
cantdi
ff
erencebet
weent
het
reat
ment
s

St
ep1

 Cal
cul
ati
onoftr
eat mentt
ot alT1,
T2…..Tt
 Cal
cul
ati
onofgrandtotal
( GT)=T1+T2+……+Tt
 Cal
cul
ati
onofcorrect
ionfactor(C.F)

St
ep2

 Cal
cul
ati
onofsum ofsquar
es

 Er
rorsum ofsquar
es(
ESS)=TSS–TrSS

St
ep3

 For
mat
ionofAnal
ysi
sofVar
iancet
abl
e

Sourceof Degreesof Sum of Mean F

v
ar i
ati
on(S.
V) f
reedom (d.
f) squares squares
(S.S) (M.S.)=
SS/d.f
.

Tr
eat
ment
s (
t-
1) TrSS
TrSS= Tr
TrMS/(t
-1) MS/EMS
 Er
ror (
N-t
) ESS
EMS=ESS/
(N
-t
)

Tot
al (
N-1) TSS

St
ep4

I
nter
pret
ati
on

 IfcalculatedF<t abl
ev alueofFf or(t-
1),(N-1)d.f.at5%level
,Fisnot
significant.H0isaccepted.Allthetreatment sareal i
ke.
 IfcalculatedF>t abFf or(t-
1),(
N- 1)
d.f.at5%l evel Fissi
gnif
icant
.F=() *
.Hoi s
rejected.
 Ii
fcal culat
edF>t abFf or(t-1)
,(N-1)d.fat1%l evelFishighl
ysignif
icant.F=(
) *
* .Hoi sreject
ed.
 IfFi ssignifi
cantorhighlysigni
f i
cant,cri
ti
cal dif
ferencebetweentreatment
meansi st
obewor kedout .

Cr
iti
caldif
fer
encebetweenanyt
wotreat
mentmeansisdefi
nedastheleast
si
gnif
icantdi
ffer
encebet
weenanytwotr
eatmentmeans,
tobeexceededbythe
di
ff
erencebetweentwotreat
mentmeanstodecl
aret
hem assi
gni
ficant
lydi
ffer
ent
.

Cri
ti
caldi
ffer
encebet
ween = Standarderrorofthe
anytwotreat
mentmeansat dif
ferencebetweenthe
5%level tr
eat mentmeansxt able
valueof‘t
’forerr
ord.f.at
5%l evel
.

Cri
ti
caldi
ffer
encebet
ween = Standarderrorofthe
anytwotreat
mentmeansat dif
ferencebetweenthe
1%level tr
eat mentmeansxt able
valueof‘t
’forerr
ord.f.at
1%l evel
.

 Cr
it
ical
dif
fer
encebet
weenTr
1andTr
2at5%(
1%)
  Cr
it
ical
dif
fer
encebet
weenTrTr
1, 3at5%(
1%)

 Af
terthis,wri
tetr
eat
mentmeansTr Tr
1, 2….
Trni
ntheascendingor
derof
magnitude.Barchar
trepresent
ati
onwi l
lbedonetoshowt hesi
gni
fi
cant
di
ff
erencesamongt hetreatments.

Not
e

 Ifwehav
eequal numberofr
epl
i
cat
ion,
i.
e,n1=n2=….nt=n,
then,
wehav
ein
step2
 Treat
mentsum ofsquar
es

 Cri
ti
caldi
ff
erencebet
weenanyt
woxtabl
e't
'forer
rord.
f.at5%or1%l
evel
tr
eatmentmeansat5%or1%level

Exer
cise

 Asetofdatainvol
vi
ng4feedstuff
sA, B,C,Dtr
iedon20chi
cksisgi
venbel
ow.
Allt
he20chicksaretr
eat
edal i
keinal
l r
espect
sexceptt
hefeedi
ng
tr
eatmentsi
sgivento5chicks.Anal
ysethedata.
 We i
ghtgai
n(ingms)ofbabychicks

A B C D

55 61 42 169

49 112 97 131

42 30 81 169
 21 89 95 85

52 63 92 154

RANDAMI
ZEDBLOCKDESI
GN(
RBD)

 Int hisdesign,wemakeuseoft hepr i

ncipleofl ocal cont r
ol wit
ht hei deaof
reduci ngthev ari
abi l
ityduet oexper iment al error.
 Firstwear ranget heexper i
ment al unitsintohomogeneousbl ockssucht hat
withint heblocks, t
heexper i
ment al unitsar easf araspossi bl
ehomogeneous
andbet weent hebl ockst hereisv ariabili
ty.
 Thi sdesigncanbemadeuseofwhent heexper iment erfindsthatt he
avai l
ableexper i
ment alunit
scanbegr oupedi ntodi f
ferenthomogeneous
blockseachcont ai
ni ngunitsasmanyast henumberoft reatment sorasa
mul tipleofthenumberoft reatment s.
 Thebl ocki
nghast obedoneont hebasi sofanyobser vablecharact er,which
i
sl ikelytohav einfluenceov ert hef actorunderst udy.
 Fore. g.inthest udyofcompar ati
v eef fectsonwei ghtgai ninchicks, the
chicksmaybegr oupedaccor dingt ost rai
ns.I fwear etoconducta
Randomi sedBl ockDesi gntot r
y‘t’treat ment s,weneed` bt'numberof
exper imentalunits, wher eb=numberofbl ocksandt=numberoft reatment s.
 Eachbl ockisar eplicati
onbyi t
selfandnumberofr epli
cation=numberof
blocks=b.Thi si sther estri
cti
oni nt hecaseofRBD.RBDi sanequi -repli
cated
desi gn.

RBDisabetterdesi
gnthanCRD,asthi
swi
l
lhav
egr
eat
erpr
eci
sionoft
heest
imat
es
andgreat
ereffi
ci
encyofthedesi
gns.

Randomi
zat
ion

 Havi
ngfor
medt henumberofblocks,experi
mentaluni
tstothedi
ff
erent
t
reat
mentswil
lbeall
ott
edindependentlyforeachbl
ockrandomlyandthe
r
esponseval
ueswil
lbetabul
atedasfollows:

Tr
eat
ment Tr
1 Tr
2 … Tr
t Tot
al

1 y11 y
12 … y
1t B1
 2 y21 y22 … y2t B2

… … … … … …

… … … … … …

Bt
b yb1 yb2 … ybt

Tot
al T1 T2 … Tt GT

STEPWI
SEPROCEDURE

 Ho:
Treat
mentmeansdonotdi
ff
erentsi
gni
fi
cant
ly

St
ep1

 Calculat
ionoftr
eatmenttotal
,i
.e.T1,T2…..Tt
 Calculat
ionofbl
ocktotal
,i.
e.B1,B2,….
Bt
 Calculat
ionofgrandtot
al(GT)=T1+T2+……+Tt=B1+B2+……+Bt=Sum of
thetreat
menttotals=Sum ofthebl ocktotal
s

St
ep2

 Cal
cul
ati
onofsum ofsquar
es
St
ep3

 For
mat
ionofAnal
ysi
sofVar
iancet
abl
e

ANOVATabl
e

Sour
ceofv
ari
ati
on d.
f. Sum ofsquar
es Meansquar
es F

Tr
eat
ment
s (
t-1) TrSS TrMS=TrSS/
(t-
1) TrMS/
EMS

(
b-1) BSS BMS=BSS/
(b-
1)
Bl
ock

(
b-1)(
t- ESS EMS=ESS/
(b-
1)(
t-
Er
ror 1) 1)

Tot
al N-
1 TSS

St
ep4

I
nter
pret
ati
on

 Ifcalculat
edF<t ableval
ueofFf or(t
-1)
,(
b- 1)
(t-
1)d.
f.at5%level
,Fisnot
signi
f i
cant.H0isaccepted.Allthetr
eatment sareali
ke.
 Ifcalculat
edF>t abFfor( t
-1),
(b-1)
(t-
1)d.f.at5%levelFissi
gnif
icant
.F=( )
*.
Hoi srejected.
 Ifcalculat
edF>t abFfor( t
-1),
(b-1)
(t-
1)d.fat1%l ev
elFishighl
y
signi
f i
cant.F=( )
**.Hoisrejected.
  I
nthelastt
wocaseswehavetocalcul
atecr
it
ical
dif
fer
encebet
weenanyt
wo
t
reat
mentmeansat5%(1%)level
.

xt
abl
eval
ueof`
t'f
orer
rord.
f.at5%(
1%)l
evel
.

 Barchar
tr epr
esent
ati
onwi
l
lbedonet
oshowt
hesi
gni
fi
cancebet
ween
tr
eatments.

ADVANTAGESANDDI
SADVANTAGESOFRBD

Adv
ant
ages

 Thisisasi mpledesignwi t
honelocalcontrolformor eefficientut
ili
sationof
theav ail
ableexperi
ment alunit
s.RBDt akesintoaccountandel iminatesthe
assignablesourceofv ariati
onamongt heexper i
mentaluni t
sbymeansof
groupi ngthemorehomogeneousuni t
stogether .
 Thisr educestheexperiment al
errorandthet estofsignif
icancebecomemor e
effi
cient.
 An ynumberoft reatment sandanynumberofr epli
cat
ionsmaybei ncluded
buteacht r
eatmentshoul dhavesamenumberofr epl
i
cations.

Di
sadv
ant
ages

 Whenthedataf
rom someexper
imental
unitsar
emissi
ngthe“Missi
ngpl
ot
t
echni
que”hastobeused.
 I
fthemissi
ngobser
vati
onsaremore,
thisdesi
gnisl
essconveni
entt
hanCRD.

Exer
cise

 Considert
heresul
tsgi
veni
nfol
lowi
ngt
abl
eforanexper
imenti
nvol
vi
ng6
tr
eatmentsin4randomi
zedbl
ocks.

Responsev
aluesf
oraRBD

Bl
ocks 1 2 3 4 5 6

1 24.
7 20.
6 27.
7 16. 16. 24.
5
2 2

2 27.
3 28.
8 22.
7 15. 17. 22.
5
0 0
 3 38.
5 39.
5 36.
8 19. 15. 26.
3
6 4

4 28.
5 31.
0 34.
9 14. 17. 22.
6
1 7

BI
OLOGI
CALASSAYS

Meani
ngofbi
oassydosager
esponsecur
ve

 Biologi cal assayorbi oassayi saf orm ofexper imentf ort heest imationoft he
pot encyofasubst anceorcompar i
ngt heef ficacyoft woormor esubst ances
bymeansoft her eact iont hatf oll
owst heirappl i
cat iont ol i
vingmat ter.
Bio–assayi sdi f
fer entf rom pur elycompar ativeexper iment .Intheusual
exper i
ment s, t
hemagni tudeofef fect sofdi fferentt r eat ment sarecompar ed
wher easi nbi o-assay s, thepot enci esoft reat ment sar ecompar ed.
 Bio-assay sar et husat ypeofexper i
ment swi tht heobj ectofcompar i
ngt he
efficacyoft woormor esubst ances, orpr epar ations, l
ikedr ugs,byusi ng
responsespr oducedbyt hem onsui tablel i
vingor gani sms.Thi stechni quei s
usedmor einphar macol ogical invest i
gationsf orcompar ingt hepot encyof
twoormor epr epar at ionsofi ndi v
idual drugs.
 Thebi o-assayi nv olv esast i
mul usappl iedt oasubj ectandt her esponseoft he
subj ectt ot hest imul us.
 Thest imul usmaybeapest icide,af ungi cide, av itami n, andsoon.The
i
nt ensi t
yoft hest imul usmaybev ariedsoast ov aryt hedosegi vent ot he
subj ect.Thedosecanbemeasur edasawei ght , av olumeoraconcent ration.
Thesubj ectmaybeani nsect ,apl ant ,abact er i
al cul tur e, etc.
 Whenast imul usi sappl iedt oasubj ecttheremaybeachangei nsome
char act eri
st i
csoft hesubj ect.Forexampl e,wei ghtoft hewhol esubj ectorof
somepar t
icul aror ganmaychange, ananal ytical valuemaychange, r
ecov ery
from sy mpt omsofadi seasemayappear ,ort hesubj ectmaydi e.Such
changesi nt hesubj ectar eknownasr esponses.Ther esponsemaybe
quant i
tativeasi nt hecaseofwei ghtorqual itat i
veasi nt hecaseofmor tal i
ty.
Themagni tudeofr esponsedependsupont hedose.
 Nor mal ly,twopr epar ationshav ingacommonef fectar et akenf orassay ing.
Oneoft hepr epar at i
onsi sofknownst rengthandi scal l
edt hest andard
prepar at i
onandt heot heri sofunknownst rengt handi scal ledtest
prepar at i
on.Theobj ect i
v eoft heassayi stoest imat et hepot encyoft het est
prepar at i
onr elativet ot hatoft hest andar dpr epar at i
on.
 Letzsandztdenot et hedosesoft hest andar dandt het estpr eparati
ons
respect i
velysucht hateachoft hem pr oducesapr e-assi gnedr esponsei n
somel i
v i
ngor gani sm.
 Thent her at i
oR=Zs/ Zti scalledt her elativepot encyoft het estpr eparat i
on.I f
 Risgr eaterthanuni ty,i
tshowst hatasmal l
erdoseoft het estpreparati
on
producesasmuchr esponseasar el
ati
velylargerdoseoft hest andard
preparationandhencet hepot encyoft het estprepar
at i
oni sgreaterthant hat
ofthest andar dpreparation.Si milarl
ywhenRi slessthan1t hepot encyoft he
testpreparationissmal l
ert hant hatofthest andardprepar ati
on.IfR=1, the
twopr eparationsar eequi -potent.
 Ana ssaywi tht wopr epar ati
onscont ainingthesameef fectivei
ngredient,
whichi sresponsiblef orther esponse, iscalledanaly
ticaldill
usionassay .An
assaywi thtwopr eparationswhi chhav eacommonef fectbutdonotcont ain
thesameef fectiv
ei ngredi ent
, i
scal edacompar
l ati
vedi l
lusionassay .

TYPESOFASSAY

 Ther
ear
et hreemaint y
pesofbio-
assay
s.Theyar e
o Directassay s
o Indirectassaysbaseduponquanti
tat
iveresponsesand
o Indirectassaysbaseduponquantalr
esponses

DI
RECTASSAY

 Indirectassay sdosesofst andar dandt estpr epar at i

onsar eadmini ster
edto
randoml yselect edident icalsubjects.Theadmi nistrati
onoft hestimul ii
s
stoppedassoonast hepr e-assignedresponsehasoccur red.I
ndir ectassays
thetolerancedoses, thedosesbel owwhi chnor esponseoccur sfort he
standar dandt estpr epar ati
ons, aremeasur eddi rectlyassoonast he
responsehasoccur red.
 Th etolerancewi llgener all
yv aryconsiderabl yfrom subj ectt osubject.Hence,
anumberoft rail
sar erequi redtoest i
mat et heav eraget oler
ance.For
obtainingtheneededdat a,thecommondesi gnsofexper i
ment sareused.
Oncedat aareobt ained, theav erageoft het olerances,zsand ztf ort he
standar dandt estpr epar ati
onsr especti
v elyarecal culat
ed.Theest imateof
therelativepot encyist henobt ainedasR=zs/ ztRi sest imatedthrough
Feill
er’
st heorem.

I
NDI
RECTASSAY

I
ndi
rectassay
sbasedonquant
it
ati
ver
esponses

 I
nt histypeofassay s,specif
ieddosesofst i
muliandthei
rresponsesare
recorded.
 Th eresponsesmaybeachangei nwei ght,
achangeinanalyti
calvalue,
time
ofsur vi
valofthesubjectandt hel
ike.
 Th erel
ationshipbet
weent hedoseandr esponsedrawnasaf requencycurv
e
i
sknownast hedoser esponsecur v
e.
 Th i
scur veisthenascertai
nedwhi chisusuall
yasigmoidalcurve.
  Next,t
hedosecorrespondingt
oagivenresponsei
sobtai
nedf r
om the
rel
ati
on.
 Suchresul
tsareobtai
nedforbothst
andardandtestpr
eparat
ions.Fi
nall
y,t
he
rel
ati
vepotencyi
sestimated.

Underi
ndi
rectassay
sbasedonquant
it
ati
ver
esponsest
her
ear
etwot
ypesassay
s.
Theyar
e,

o Par
all
ell
i
neassays,
o Sl
ope–rati
oassays.

Par
all
ell
ineassay
s

 Thepar all
elli
neassay sar et hosei nwhi chtherelati
onshi pbetweenthe
quant i
tativeresponseandl ogdosei slinear.Thel i
nesf orthestandardand
testprepar ati
onsshal lbepar all
el.
 Ap ar
allell
ineassayi nwhi cht hestandar dandtestpr eparationshavean
equal numberofdosesandanequal numberofsubj ectsforeachdosei s
calledasy mmet ri
calpar all
ellineassay .Otherwise,iti
scal l
edan
asy mmet r
icalparall
elli
neassay .
 Inasy mmet ricalparal
lel l
i
neassayt herear ekdosesofeachofst andardand
testprepar ati
ons.Inall therear e2kdosesi nthisassay .Toeachoft hedoses
'
n' subjectsar eall
ottedatr andom.Hence, i
tiscalleda2k–poi ntsymmetri
cal
par al
l
el li
neassay .Theconst antkmaybe2, 3,4,etc.When, k=2,wehav e4
pointpar all
ellineassay .Similarly
,wehav e6poi ntparallell
ineassay,8–
pointpar all
ellineassay ,etc.Amongt hese, 4–poi ntand6poi ntassaysare
usedv er ycommonl y.Themostpopul ardesi gnisthe4- pointassay.

Sl
oper
ati
oassay
s

 Whent heresponseisli
nearl
yrel
atedtothedose,xrai
sedtopoweri
,the
assayisknownassl operati
oassay.Thev al
ueofii
stakenas1whichis
adequateandv er
ycommonl yusedconst ant
.Theequat
ionsf
orthel
inesof
thestandardandtestprepar
ati
onswillbe,

y
s=a+bsxs y
t=a+btxt

 Thet wolinesconvergeatzerodose.Ther elat

ivepotencyis,
theref
ore,
esti
mat edfrom t
heratioofthesl opes(regressi
oncoef f
ici
ents)ofthef
it
ted
l
ines.Hencet henamesl operati
oassay .
 Thedose–r esponserelati
onshipsforthet wopr epar
ati
onsmaybe
expressedasamul t
ipl
elinearequations,

y
s=a+bsxs+a+btxt

 Theesti
matesofbsandbtar
ebsandbt,
respect
ivel
y.Theest
imat
eoft
he
rel
ati
vepot
encypist
hendefi
nedas,
  Asinthecaseofparal
lell
ineassays,t
hesymmetri
caldesi
gnsareidealfort
he
sl
oper at
ioassay
s.Hence,weshal l
consi
derher
eonlyassayswith(2k+1)
doses.Wemayhav e3–poi nt,5–point,
7–point,et
c.,
assays.The3-point
assaysisthemosteff
ici
entdesign.

QUANTALRESPONSEASSAY

 Inmanybi ological assay sther esponsesar equal i

tati
veinnat ure,forexampl e,
i
nt heassayofi nsecticidest heresponsei smor tal
ityofinsects.Such
qualit
at i
veresponsesar eal soknownasal l–or–noner esponses.The
assay sinwhi cht heresponsesar equalitativeareknownasquant alresponse
assay s.
 Inthequant al r
esponseassay ,thesubjecti sgivenapr edet er
mi neddoseof
thepr eparationundert estandi sobser vedt oseewhet herornotaspeci fi
ed
responseoccur s.
 Thus, quantal r
esponseassay sareclosel yrelatedt odir
ectassay s.Int his
typeofassay , t
hest rengthofapr eparati
oni scharactersiedbyt hemedi an
toler
anceort hedoset hatinduces50%r esponses.
 Ifther esponsei smor tali
tyitiscalledmedi anl ethaldoseandi sdenot edby
LD50.
 Ifther esponsei snotmor t
al i
ty,i
tmaybecal edmedi
l anef f
ectivedose( ED50)
,
medi anknockdowndose( KD50),medianant i
feedingdose( AD50)andt hel i
ke.
 Themostcommonl yusedmeasur ementi sLD50.Ther ati
oLD50/ED50i scal l
ed
therapeut i
cindex.

METHODSOFESTI
MATI
NGLD50

Whenthenumberofsubj
ectsusedisrelat
ivel
ysmallandthedosesar
efai
rl
ycl
ose,
LD50maybeesti
matedusingDragstedt–Behr en’
smethod.Theest
imat
eisgi
ven
by

wher
e,

 A=dosecorr
espondi
ngtoaper
cent
ageofmor t
ali
tyi
mmediatel
ybel
ow50%
mor
tal
it
y,
 a=obser
vedmortal
it
y(%)i
mmediat
elybel
ow50%mor tal
i
ty,and
 b=obser
vedmortal
it
y(%)i
mmediat
elyabove50%mortal
it
y
 SPEARMAN-KARBERMETHOD

 Anothersimpl
emet hodofest i
mationofLD50isSpearman–Kar bermet hod.
Iti
susedwhent helogdosesar eequall
yspaced.Supposethatt
hel ogdoses
aredenotedbyx1,
x2,….
.xk.I
fthelogdosesareequispaced,
thenxi+1–xi=dfor
all
i.
 TheLD50isest
imatedas

Wher
e,

xk=thehighestl
ogdose,
t
h
pi=proport
ionofresponsef
ori dose.