REVIEW
of
FINANCE -
Apr. 2018
43
Binary Logistic Regression Analysis
TABLE 3. RELIABILTY TEST FOR BINARY LOGISTIC REGRESSION
OMNIBUS TESTS OF MODEL COEFFICIENTS
Step 1 Chi-square
Df
Sig.
Step
71.072
8
.000
Block
71.072
8
.000
Model
71.072
8
.000
HOSMER AND LEMESHOW TEST
Step
Chi-square
df
Sig.
1
11.154
8
.193
MODEL SUMMARY
Step
-2 Log
likelihood
Cox & Snell
R Square
Nagelkerke
R Square
1
281.382a
.470
.627
a. Estimation terminated at iteration number 6 because
parameter estimates changed by less than .001.
Table 3 shows that the Omnibus Test of Model
coeffiencent with Sig.< 0.05 and Hosmer and
Lemeshow Test with Sig. > 0.05 together with the
-2 Log Likelihood is 281.382a, which defines that
the model is fit to the data.
TABLE 4. CLASSIFICATION TABLEA
Observed Predicted Decision Percentage
Correct
Step 1
No
Yes
Decision
No
165
34 82.9
Yes
28
149 84.2
Overall
Percentage
83.5
a. The cut value is .500
Source: Authors
The table 4 shows that 165 cases are correctly
predicted be “No” among the 199 cases observed,
approximately 82.9%; and 149 cases are correctly
predicted be “Yes” among the 177 case observed,
approximately 84.2%. Therefore, the overall
percentage of themodel is approximated to be 83.5%.
From these results, all five parameters have
added significantly to the model (Sig. < 0.05).
There should be no high intercorrelations
(multicollinearity) among the predictors. This
can be assessed by a correlation matrix among
the predictors because correlation coefficients
among independent variables amount to less
than 0.8.
TABLE 5. VARIABLES IN THE EQUATION
B S.E.
Wald Df Sig. Exp(B)
Step
1a
PC 1.270 .196 41.847 1 .000 3.561
JC .455 .172 7.030 1 .008 1.576
UC .708 .241 8.631 1 .003 2.029
FP 1.243 .276 20.275 1 .000 3.465
CD 1.944 .302 41.444 1 .000 6.984
Constant
-19.952 2.102 90.057 1 .000 .000
a. Variable(s) entered on step 1: PC,JC,UC,FP,CD
CORRELATION MATRIX
PC
JC
UC
FP CD
Step 1
PC 1.000 -.054 -.113 .074 .279
JC -.054 1.000 .028 -.117 .043
UC -.113 .028 1.000 -.063 .153
FP .074 -.117 -.063 1.000 .126
CD .279 .043 .153 .126 1.000
Source: Authors
Linear regression predicts the value that Y
takes.
E(
Y
X
i
) =
P
1−P
= e
(-19.952 + 1.27PC +0.455JC + 0.708UC + 1.243FP + 1.944CD)
P: probability of Y occuring
Given that the other variables don’t change at
the same time, everytime Personal Characteristics
increases by 1 unit, the probability of the Decision
to choose Accounting as a major increases by 1.27
times; Job characteristics increases by 1 unit will
make the probability of the Decision to choose
Accounting as a major increase by 0.455 times;
University characteristics increase by 1 unit will
cause the probability of the Decision to choose
Accounting as a major to increase by 0.708 times;
Family members and peers increases by 1 unit
will also drive up the probability of the Decision
to choose Accounting as major by 1.243 times;
and finally Communication and Labour demand
increases by 1 unit will increase the probability
of the Decision to choose Accounting as major by
1.944 times.
=
݁
−19.952 + 1.27PC +0.455JC + 0.708UC + 1.243FP + 1.944CD )
1 + ݁
−19.952 + 1.27PC +0.455JC + 0.708UC + 1.243FP + 1.944CD )
P(
Y
X
i
)
Xi: Predicts the probability of Y at Xi
Assumption that a person has the average
value of PC = 3.7, JC=3.8, UC=3.4, FP=4.0, CD=4.0,
so the probability of Y is:
