Engineering Research Models Evaluation Methods

International Journal on Recent and Innovation Trends in Computing and Communication
Volume: 4 Issue: 2
ISSN: 2321-8169
158 - 161
________________________________________________________________________________________________________
Engineering Research Models Evaluation Methods

Awari Mahesh Babu
Professor
Civil Engineering Department,
Tirumala Engineering College,
Hyderabad.TS. India.
E-mail: [email protected]
Abstract: The objective of several records analysis is in the direction of extorts since unrefined in sequence the precise evaluation. Individual of
the mainly significant as well as frequent difficulty regarding but here is statistical correlation among a retort unpredictable Y and instructive
unpredictable Xi. An alternative in the direction of response this problem is just before take up regression analysis in order to model its
correlation. There are different kinds of regression analysis. The category of the regression model depends scheduled the category of the
distribution of Y. In modeling it is to anticipate the result Y stand scheduled standards of a set of unpredictable Xi. In research problems and
models regression analysis and evaluation of methods are very important In this paper it is focused in regression and evaluation methods for
validation of models.
Keywords: Regression analysis, models, variance, covariance, evaluation
__________________________________________________*****_________________________________________________
I.
INTRODUCTION
In statistical modeling, regression analysis is a

statistical method for estimate the relations between
unpredictable. It contains diverse methods on behalf of
modeling moreover evaluate a number of unpredictable,
Further overtly, regression analysis facilitate one identify
how the attribute assessment of the reliant unpredictable
changes while several individual of the autonomous
unpredictable is mottled, whereas the other autonomous
unpredictable are apprehended predestined In regression
analysis, it is besides of significance toward distinguish the
deviation of the reliant unpredictable about the regression
role which know how to be described with a probability
distribution. Regression analysis and evaluation methods
experimental values of SPEED and Stopping Sight Distance
(SSD) are indentified.
II.
REGRESSION ANALYSIS
Various regressions give an equation that calculate

one unpredictable from two or more autonomous
unpredictable,
Y = 0 + 1X1 +
2X2 ..---------- (1)
Important steps concerned in performing a regression
analysis
Step-1: Creation of the regression model
The construction of an instructive model is a fundamental
step in the regression analysis. It has to be described
through position to the action theory of the intervention. It
is expected that a number of category of unpredictable exist.
An unpredictable might also correspond to an observable
attribute or an unobservable one.
The model may imagine that a particular unpredictable

evolves in a linear, logarithmic, exponential or other way.
All the instructive models are constructed on the base of a
model, such as the subsequent, for linear regression:
Y = 0 + 1X1 + 2X2 + . + kXk , where -------- (2)
Y is the differ that the program is principally theoretical to
make
X1-k are autonomous unpredictable likely to explain the
change.0-k are constants.
Step-2: Construction of a model
To apply multiple regressions, a large model is generally
essential.
Step-3: Records collection
Consistent information have to be collected, any from a
monitoring scheme, from a questionnaire survey or from
field survey.
Step-4: Computation of coefficients
Coefficients know how to be premeditated rather easily,
using statistical software that is both reasonable and
available to PC users.
Step-5: Valuation of the model
The model plan to elucidate as much of the variability of
the experimental changes as possible. Assessment can be
done from graphical and computational analysis and
significance tests. The percentage of the variability in the y
variable can be explained by the x variable.
Flowchart showing the step by step procedure of
Regression analysis is presented in Figure 1
158
IJRITCC | February 2016, Available @ http://www.ijritcc.org
________________________________________________________________________________________________________

Volume: 4 Issue: 2
ISSN: 2321-8169
158 - 161
________________________________________________________________________________________________________
SSR is defined as
= =1 2 --------------- (4)
Step 1. Creation of the model
Y = f (X1, X2, ..., Xn)
Where is the response vector
SST, which is furthermore called as squares about the mean,
is defined as
= =1 2 --------------- (5)
Step 2. Creation of a model
Where SST = SSR+SSE
Therefore R can be expressed as
For illustration, in the course of
a
opinion poll
survey
Step 3. Compilation of
information
Step 4. Computation of the

coefficients of the correlation
form
Step 5. Analysis of the
coefficients acquired
Importance of parameterisation
study of dregs
Step 6. Overview to the entire

presumption
Fig 1 Flow Chart for Regression analysis

III.
EVALUATION OF MODELS
For analysis of the data it has to be fitted with one or

more models. The analysis deals numerical fit analysis,
graphical fit analysis and significance tests. The graphical
measures deals with the residuals and prediction bounds and
the numerical measures deals with the goodness of fit
statistics as given below
1. Confidence and Prediction bounds
Confidence bounds pro the integral coefficients with
prediction bound for new observations or the integral task
know how to be considered. The confidence bounds are
numerical which are displayed through entity polynomial
coefficients, while the prediction bounds are displayed
graphically. Confidence and prediction bounds delineate the
subordinate and superior values of the connected period, and
describe the size of the interval.
3. Sum of squares due to errors (SSE)
The above statistic measures the total deviation of the
response values from the fit to the response values. A value
closer to zero indicates a better fit. Mathematically it is
represented as
= =1 2 ------------- (3)
Where is the response value and is the predicted value
4. R
These statistics dealings how successful the fit is in
explaining the variation of the data. It is the square of the
correlation between the response values and the predicted
response values.
= 1
-------------- (6)
The coefficient of determination, R, is constructive

since it gives the quantity of the discrepancy (fluctuation) of
one unpredictable that is predictable from the other variable.
It is a measure that allows us to determine how certain one
can be in making predictions from a certain model/graph.
The coefficient of determination is the ratio of the explained
variation to the total variation.
The coefficient of determination is such that 0 < R2 < 1,
and denotes the strength of the linear association between x
and y.
5. Degree of freedom (DFE)
The residuals degree of freedom is distinct as the
quantity of response values n less the quantity of fitted
coefficients m predictable from the response values.
= --------------- (7)
v designate the quantity of independent pieces of
information on which the approximation is based and is
known as Degree of Freedom (DFE) of an estimate. A lesser
amount of the degree of freedom less is the complication
and degree of freedom also helps to evade the over fitting of
the data.
6. Root Mean Square Error RMSE. It is also known as
Standard Error
This statistic is also identified as the standard error of
the regression
= = ---------- (8)
Where MSE is the mean square error or the residual mean
square
=
-------------
(9)
Standard Error (RMSE) value closer to 0 indicates a better
fit.
7. Multiple correlation coefficients R
It is the square root of R2
Linear correlation coefficient,
r = (xy) sqrt [( x2 )
* ( y2 )]---(10)
The quantity r, called the linear correlation coefficient,
measures the potency and the direction of a linear
correlation between two variables. The value of r is such
that -1 < r < +1. The + and signs are used for positive
linear correlations and negative linear correlations,
159
________________________________________________________________________________________________________

Volume: 4 Issue: 2
ISSN: 2321-8169
158 - 161
________________________________________________________________________________________________________
The Strength and Significance coefficients:
respectively. ve sign indicates that one variable increases
and other will decreases and vice versa.
MCC=1, SSE =2.52, R= 1, SE= NaN, DFE =0
IV.
REGRESSION ANALYSIS AND EVALUATION

OF MODELS
For case study transportation engineering problem is

premeditated. The following table values are experimental
values of speed and stopping sight distance
Table 1
Nodes/
Criterions
SPEED
SSD
32.006
49.109
33.541
50.522
32.368
41.615
32.314
44.423
28.007
40.061
Table 2 Comparison of risk parameters obtained from

above analysis
Polynomi
al degree
3rd degree polynomial
4th degree polynomial
Paramete
r
SS
E
RMS
E
DF
E
SS
E
RMS
E
DF
E
SSD
0.7
6
0.9
6
0.87
2.5
2
NaN
Good fit
polynomi
al
3rd
iv) Comparison of parameters to get good fit graphical

analysis for SPEED and SSD
Computation of good fit for regression models

Computation of good fit for SPEED and SSD (Experimental
values)
Linear model Poly3
f(x) = 1 3 +n2 2 +n3x+n4-------(11)
Fig 2 Curves for good fit of SPEED and SSD
Where x is normalize by mean 45.15 and std 4.568.

Coefficients (with 95% confidence bounds):
n1 = 3.864 (-14.47, 22.19), n2 = -1.384 (-11.85, 9.083)
n3 = -2.596 (-23.56, 18.37), n4 =
32.42 (22.42, 42.42)
The above information revealed for absolute ideals of

Speed and Stopping sight distance. In Residual graph,
residual data of 3rd degree polynomial data become visible
arbitrarily which suggests the model fits the data well.
The Strength and Significance coefficients:

Multiple correlation coefficient -MCC= 0.98, Sum of
squares
due
to
errors
SSE
0.76,coefficient
of
determination R= 0.96, Standard Error-SE = 0.87, Degree

of freedom (DFE)=1
Linear model Poly4:
f(x)=n1 4 + 2 3 +n3 2 +n4x+n5--------(12)
Where x is normalized by mean 45.15 and std
4.568,Coefficients:
n1 = -2.039, n2 = 4.357, n3 = 1.555, n4 = -3.068
n5 =
31.81
Fig 2.2 Prediction bounds for fit using poly3

160
________________________________________________________________________________________________________

Volume: 4 Issue: 2
ISSN: 2321-8169
158 - 161
________________________________________________________________________________________________________
V.
CONLUSION
It is observed that 4th degree polynomial model gives R
significance identical to 1 however it more than fits the data
and amplify the computational intricacy. From graphical
analysis it can be measured that 3rd degree polynomial fits
the model and it is also observed that 3rd degree polynomial
SSE is less than 4th degree polynomial which indicates better
fit. The deterioration analysis contrivance is advanced tools
that be capable of categorize how unlike variables in a
method are linked. The regression tool will inform if one or
multiple variables are unified among a method output. This
in order know how to categorize where in the method direct
is desired or what factors is the unsurpassed first point for a
method upgrading task.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
M. H. Kutner, C. J. Nachtsheim, and J. Neter (2004),

"Applied Linear Regression Models", 4th ed., McGrawHill/Irwin, Boston (p. 25)
N. Ravishankar and D. K. Dey (2002), "A First Course in
Linear Model Theory", Chapman and Hall/CRC, Boca Raton
(p. 101)
Steel, R.G.D, and Torrie, J. H., Principles and Procedures of
Statistics with Special Reference to the Biological Sciences.,
McGraw Hill, 1960, page 288.
Chiang, C.L, (2003) Statistical methods of analysis, World
Scientific. ISBN 981-238-310-7 - page 274 section 9.7.4
"interpolation vs extrapolation"
Good, P. I.; Hardin, J. W. (2009). Common Errors in
Statistics (And How to Avoid Them) (3rd ed.). Hoboken, New
Jersey: Wiley. p. 211. ISBN 978-0-470-45798-6.
Tofallis, C. (2009). "Least Squares Percentage Regression".
Journal of Modern Applied Statistical Methods 7: 526534.
doi:10.2139/ssrn.1406472.
YangJing Long (2009). "Human age estimation by metric
learning for regression problems" (PDF). Proc. International
Conference on Computer Analysis of Images and Patterns:
7482.
Rothman KJ, Greenland S. Modern Epidemiology. 2nd ed.
Lippincot-Raven; 1998.
Altman DG. Practical Statistics for Medical Research.
Chapman & Hall/CRC; 1991. .
Draper NR, Smith H. Applied Regression Analysis. Wiley
Series in Probability and Statistics; 1998. Applied Regression
Analysis.
Munro BH. Statistical Methods for Health Care Research. 5th
ed. Lippincott Williams & Wilkins; 2005.
161
________________________________________________________________________________________________________

Engineering Research Models Evaluation Methods

Uploaded by

Copyright:

Available Formats

Engineering Research Models Evaluation Methods

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Engineering Research Models Evaluation Methods

Uploaded by

Copyright:

Available Formats

International Journal on Recent and Innovation Trends in Computing and Communication