Muhammad Bulkhan Bin Ismail

EGE 080043

Master Applied Statistics

For

ESGC 6114

Associate Professor Dr. Rohana Jani

Tutorial 7

Question 2

a) Perform a multiple linear regression analysis using the above variables (Full Model). State the equation.

Y = β0 + β1X1 + β2X2 + β3X3 + β4D1 + β5D2 + β6D3 + β7D4 + β8D5 + β9D6 + β10D7 + β11D8 + β12D9 + β13D10 + β14D11

Where,

X1 : Age of Employee (Data Collected 1995, 95-Yrborn (Year employee born)) AgeEmp
X2 : Number of years of work experience at another bank prior to working at Fifth National (YrsPrior)
X3 : Years of experience with The Fifth National bank (YrsExper)
D1 : 1 for finished high school, 0 for otherwise
D2 : 1 for finished some college course, 0 for otherwise
D3 : 1 for obtained a bachelor’s, 0 for otherwise
D4 : 1 for took some graduate courses, 0 for otherwise
D5 : 1 for job level 1, 0 for otherwise
D6 : 1 for job level 2, 0 for otherwise
D7 : 1 for job level 3, 0 for otherwise
D8 : 1 for job level 4, 0 for otherwise
D9 : 1 for job level 5, 0 for otherwise
D10 : 1 for male, 0 for otherwise
D11 : 1 for employee current job is computer-related, 0 for otherwise

Coefficients(a)

Model Unstandardized Standardized

Coefficients Coefficients t Sig.

B Std. Error Beta B Std. Error

1 (Constant) 53.659 3.524 15.226 .000
AgeEmp -.009 .058 -.008 -.155 .877
YrsPrior .168 .140 .047 1.194 .234
YrsExper .516 .098 .320 5.262 .000
Edu1 -2.691 1.621 -.091 -1.660 .099
Edu2 -3.176 1.561 -.106 -2.035 .043
Edu3 -2.163 1.170 -.089 -1.849 .066
Edu4 -2.406 2.166 -.041 -1.111 .268
Job1 -23.832 2.800 -.962 -8.512 .000
Job2 -22.268 2.742 -.796 -8.121 .000
Job3 -18.613 2.625 -.671 -7.091 .000
Job4 -15.238 2.456 -.463 -6.205 .000
Job5 -10.173 2.370 -.273 -4.293 .000
Gender 2.554 1.012 .107 2.524 .012
Pc_Job 4.923 1.474 .126 3.340 .001
a Dependent Variable: Salary
b)
i. Determine the tolerance limit and VIF for each predictor variable. Is there reason to suspect the
existence of multicollinearity

Coefficients(a)

Model Unstandardized Standardized

Coefficients Coefficients Collinearity Statistics

B Std. Error Beta t Sig. Tolerance VIF

1 (Constant) 53.659 3.524 15.226 .000
AgeEmp -.009 .058 -.008 -.155 .877 .435 2.300
YrsPrior .168 .140 .047 1.194 .234 .795 1.258
YrsExper .516 .098 .320 5.262 .000 .329 3.042
Edu1 -2.691 1.621 -.091 -1.660 .099 .408 2.452
Edu2 -3.176 1.561 -.106 -2.035 .043 .450 2.224
Edu3 -2.163 1.170 -.089 -1.849 .066 .531 1.884
Edu4 -2.406 2.166 -.041 -1.111 .268 .884 1.131
Job1 -23.832 2.800 -.962 -8.512 .000 .095 10.491
Job2 -22.268 2.742 -.796 -8.121 .000 .127 7.901
Job3 -18.613 2.625 -.671 -7.091 .000 .136 7.367
Job4 -15.238 2.456 -.463 -6.205 .000 .218 4.580
Job5 -10.173 2.370 -.273 -4.293 .000 .301 3.324
Gender 2.554 1.012 .107 2.524 .012 .681 1.469
Pc_Job 4.923 1.474 .126 3.340 .001 .851 1.176
a Dependent Variable: Salary

By referring the above SPSS output, variables AgeEmp, YrsExp, Edu1, Edu2, Job1, Job2, Job3, Job4 and Job5 have
tolerance level lower than 0.5, which indicate multicollinearity problem exist. For the other variables having
tolerance higher than 0.5 and VIF close to 1 indicate that these variables are not correlated.
ii. Using the same data set perform a stepwise regression analysis.

Coefficients(a)
Unstandardized Standardized
Model Coefficients Coefficients Collinearity Statistics
B Std. Error Beta t Sig. Tolerance VIF
1 (Constant) 30.330 1.055 28.761 .000
YrsExper .992 .088 .616 11.212 .000 1.000 1.000
2 (Constant) 33.599 1.050 31.993 .000
YrsExper .916 .080 .569 11.454 .000 .983 1.018
Job1 -8.808 1.231 -.355 -7.155 .000 .983 1.018
3 (Constant) 31.023 1.054 29.429 .000
YrsExper .918 .074 .570 12.458 .000 .983 1.018
Job1 -7.598 1.151 -.307 -6.603 .000 .954 1.048
Gender 6.766 1.102 .283 6.140 .000 .970 1.031
4 (Constant) 34.296 1.149 29.841 .000
YrsExper .815 .071 .506 11.430 .000 .916 1.092
Job1 -9.930 1.155 -.401 -8.597 .000 .827 1.210
Gender 6.293 1.033 .263 6.090 .000 .964 1.038
Job2 -7.188 1.300 -.257 -5.529 .000 .831 1.203
5 (Constant) 39.094 1.474 26.514 .000
YrsExper .703 .072 .436 9.817 .000 .819 1.221
Job1 -13.466 1.318 -.543 -10.220 .000 .572 1.748
Gender 4.612 1.040 .193 4.433 .000 .856 1.168
Job2 -10.692 1.431 -.382 -7.473 .000 .618 1.618
Job3 -6.932 1.433 -.250 -4.836 .000 .605 1.652
6 (Constant) 45.111 1.822 24.754 .000
YrsExper .594 .071 .369 8.402 .000 .746 1.341
Job1 -18.338 1.566 -.740 -11.707 .000 .360 2.777
Gender 3.339 1.012 .139 3.298 .001 .804 1.243
Job2 -15.570 1.653 -.557 -9.419 .000 .412 2.428
Job3 -11.856 1.660 -.428 -7.141 .000 .401 2.492
Job4 -8.512 1.666 -.259 -5.109 .000 .561 1.783
7 (Constant) 56.056 2.864 19.573 .000
YrsExper .406 .078 .252 5.222 .000 .556 1.799
Job1 -27.450 2.413 -1.108 -11.376 .000 .137 7.310
Gender 1.927 1.005 .080 1.916 .057 .735 1.360
Job2 -24.783 2.481 -.886 -9.990 .000 .165 6.065
Job3 -21.030 2.479 -.758 -8.484 .000 .162 6.163
Job4 -16.926 2.362 -.514 -7.165 .000 .251 3.976
Job5 -11.335 2.364 -.304 -4.796 .000 .322 3.102
8 (Constant) 54.366 2.809 19.353 .000
YrsExper .442 .076 .274 5.821 .000 .547 1.827
Job1 -26.585 2.348 -1.073 -11.323 .000 .136 7.380
Gender 2.786 1.000 .116 2.787 .006 .697 1.434
Job2 -24.500 2.403 -.876 -10.194 .000 .165 6.071
Job3 -20.397 2.406 -.736 -8.477 .000 .161 6.193
Job4 -16.524 2.290 -.502 -7.216 .000 .251 3.985
Job5 -10.532 2.299 -.283 -4.582 .000 .320 3.128
Pc_Job 5.376 1.423 .138 3.779 .000 .913 1.096
a Dependent Variable: Salary
iii. State the equation

X3 : Years of experience with The Fifth National bank (YrsExper)

D5 : 1 for job level 1, 0 for otherwise
D6 : 1 for job level 2, 0 for otherwise
D7 : 1 for job level 3, 0 for otherwise
D8 : 1 for job level 4, 0 for otherwise
D9 : 1 for job level 5, 0 for otherwise
D10 : 1 for male, 0 for otherwise
D11 : 1 for employee current job is computer-related, 0 for otherwise

Y = 54.366 + 0.442X3 + 2.786D10 + 5.37611 - 26.585D5 - 24.500D6 - 20.397D7 - 16.524D8 - 10.532D9

iv. Perform a residual analysis and determine the adequacy of the model

The above graph, does not show increasing or decreasing in residual that indicates the presence of
homoscedasticity
The histogram and normal p-p plot look normal we can conclude that the homoscedasticity and normality
assumption are satisfied.
From the above partial regression plot, we can see that the overall pattern is linear. Therefore on the basis of a residual
analysis, the models appear adequate.