A O A C O ffic ia l M e t h o d s of A n a l y s is (1990) M ilk 817

as in 925.21 and analyze in triplicate. Compare averages of be performed outside instrument. To perform this calcn on cal­
second and third instrument readings with std values and fol­ ibration data that are collected on successive days or weeks,
low directions in operating manual for making required ad­ it is necessary to record regression ests as well as final result.
justments to calibration. This alternative, preferred method has been demonstrated to
Alternatively, for better accuracy, use simple linear regres­ be more accurate.
sion equation that relates instrument ests and std ref. values, Note: For products that are fortified or dild, both original
using latter as dependent variable, to correct these ests. In ear­ multiple regression and simple regression ealens must be based
lier instruments, which do not have this capability in instru­ on regression that forces calibration line thru origin. First calcn
ment software, this calcn must be performed outside instru­ procedure in this sec., par. 1, cannot be used successfully with
ment. To perform this calcn on calibration data that are collected these products.
on successive days or weeks, it is necessary to record instru­
K. Computer-Based Calibration and C ontrol
ment ests as well as final results. Purpose of this alternative,
preferred procedure is to avoid adjusting instrument slope con­ Program computer to:
trols. (7) Calc. S2' = Sj + [(S2 — S |)/PF], where S2' is purge-
corrected signal; S2 is signal at any instrument channel; S, is
G. D eterm ination same signal for previous sample; and PF is predicted purge
Prep, sample as in 925.21. Operate instrument according to fraction = purging efficiency/100 (see 972.16C). Apply cor­
manufacturer's instructions. rection to all signals received from instrument.
(2) Calc. (SCS2')/1 0 , where CS2' is purge-corrected con­
trol milk signal. Sum control milk signals at each channel for
III. Lactose 10 control samples, and average. At time of calibration, store
and designate as required control milk samples.
H. Calibration (i) Calc. S = S2' + CSE, where S is purge- and drift-cor­
Proceed as in 972.161). par. 1, substituting lactose soln (dis­ rected signal; CSE is control signal error = (required control
solve 50.00 g lactose. H20 and 0.25 g HgCl2 in H20 and dil. signal av.) —(detd control signal av.) for any check on control
to 1 L with H20 at 20°) for homogenized cream to prep, mixts milk. Apply value to all except control milks.
of known relative concns ranging from 0 to required max. for (4) Calc, ests from purge- and drift-corrected signals using
lactose. Det. lactose for series of ^ 8 preanalyzed (896.01 B or calibration signals, and store. Equation type is Est = A + B
930.28) milk samples having range of lactose content approx, (main equation). Initially, A = 0 and B = 1.
that of population of milks to be analyzed. (5) Calc, calibration equations and store.
Prep, samples as in 925.21 and analyze in triplicate. Use (6) Store random sample signals, ests, and std method ref.
averages of second and third values for each sample in estg results.
calibration requirements. Adjust instrument controls, as di­ (7) Calc, and store mean difference, std dev. of difference,
rected in operating manual, to make 2L = XL', where L = limits for mean difference, and population std dev. of differ­
instrument readings for lactose and V = values from ref. ence for statistical quality control.
method. (8) Store results for checks on purging efficiency tests, ho­
Alternatively, for better accuracy, multiply instrument ests mogenization efficiency tests, linearity tests, and repeatability
by S L '/^ L to obtain final results. In earlier instruments, which tests.
do not have this capability in instrument software, this calcn
must be performed outside instrument. To perform this calcn L. Instrum ent C ontrols
on calibration data that are collected on successive days or For instrument with secondary slope controls, operate in un­
weeks, it is necessary to record instrument ests as well as final corrected mode. For instruments without secondary slope con­
results. Purpose of this alternative, preferred procedure is to trols, reduce interference correction coefficients to 0.
avoid adjusting instrument slope controls.
M. C ontrol Milk
I. D eterm ination Prep, adequate supply of pasteurized, homogenized, pre­
Prep, sample as in 925.21. Operate instrument according to served milks with % fat about that of population av. Transfer
manufacturer's instructions. amts for single tests to sample containers and keep refriger­
ated. Zero-adjust instrument channels. Heat 11 samples to 40°,
and pump each thru instrument, collecting signals. Average
IV. Solids (T otal) last 10 results for each signal to calc, required signal. Record
required control milk signals to 4 decimals to prevent rounding
J. Calibration
errors in continuous control situation.
Use std method 925.23A to det. % total solids for series of
>8 milks. Det. % fat, % protein, and % lactose. Calc, mean N. Setting Prim ary Slope C ontrols
difference for total solids (TS) — F — P — L — a, where F , Det. % fat, protein, and lactose on series of >8 preanalvzed
P . and L are ests of fat, protein, and lactose, resp. For routine (e.g., 905.02, 920.105, 896.01 A. or 984.15) milks of type to
control of calibration, analyze addnl series of milks and adjust be analyzed by instrument. Warm samples to 40°, mix thoroly,
value for mean difference in accordance with accumulated data. and pump thru instrument, collecting and correcting signals for
purge and drift. For each signal, calc, regression equation of
Calc. % total solids = a + F + P + L
type S = slF + b/5 + cL, where S is signal and F, P. and L
Alternatively, calibrate by multiple regression to calc, equa­ are results for std. At fat channel, adjust slope amplification
tion for estn of either TS or SNF as function of fat, protein, by 1/a. At protein and lactose channels, adjust by 1/b and 1/
and lactose uncorrected signals. For rccalibration, use simple c, resp. Check success of attempt to obtain coefficient of 1
linear regression equation which relates regression ests and std for main components by repeating test. Readjust slope ampli­
ref. values, using latter as dependent variable, to correct these fications if necessary. Lock slope controls and do not change
ests and obtain final result. In earlier instruments, which do settings thereafter. Warm 11 samples of control milk and re
not have this capability in instrument software, this calcn must det. required control milk values as indicated above.