Correlation Analysis
Correlation analysis was conducted to establish the relationship between the independent and
dependent variables. The correlation matrix is presented in Table 4.10.
Table Correlation Matrix
Employee Career Compens Performance Recruit
Performance Development ation appraisals ment
Employee
Performance 1.000
Career
Development .613** 1.000
0.000
Compensation .625** .657** 1.000
0.000
Performance
appraisals .569**
0.000
.615** .661** 1.000
0.000 0.000 0.000
Recruitment .640** .686** .676** .624** 1.000
0.000 0.000 0.000 0.000
Source: Field Survey Data (2022)
The results in Table 4.11 revealed that Career Development and employee performance of public
hospitals in Kajiado County is positively and significantly related (r= .613**, p=0.000). The
results further indicated that Compensation and employee performance of public hospitals in
Kajiado County is positively and significantly related (r= .783**, p=0.000). Performance
appraisals and employee performance of public hospitals in Kajiado County is positively and
significantly related (r= .569**, p=0.000). Lastly, results showed that Recruitment and employee
performance of public hospitals in Kajiado County is positively and significantly related (r=
.640**, p=0.000). This implies that an increase in career development, compensation,
performance appraisals and recruitment leads to an increase on employee performance of public
hospitals since the coefficients are positively related.
Diagnostic Tests
The diagnostic tests conducted included Multicollinearity Test, Test for Heteroscedasticity and
Normality Test.
Multicollinearity Test
Multicollinearity test was conducted to determine if two or more of the predictor
(independent) variables in the regression model was highly correlated. Variance inflation factor
(VIF) were used to test multicollinearity and VIF of below 10 indicated acceptable limits. If the
VIF value of exploratory variables are greater than 10, then variables were regarded as highly
collinear.
Table Multicollinearity Test Using Tolerance and VIF
Collinearity Statistics
Tolerance VIF
(Constant)
Career Development 0.441 2.270
Compensation 0.419 2.388
Performance appraisals 0.482 2.073
Recruitment 0.421 2.377
Source: Field Survey Data (2022)
From the findings above all the variables had tolerance values >0.2 and VIF values <10 as shown
in Table 4.11 and thus according to Myres (2015) who indicated that where VIF ≥ 10 indicate
presence of Multicollinearity, there was no multicollinearity among the independent variables.
Test for Heteroscedasticity
Heteroscedasticity is the circumstance in which the variability of a variable is unequal across the
range of values of a second variable that predicts it. Running a regression model without
accounting for heteroscedasticity would lead to unbiased parameter estimates. To test for
heteroscedasticity, the Breusch-Pagan/Godfrey test was used. Heteroscedasticity test was run
using Breusch-Pagan / Cook-Weisberg test in order to test whether the error terms are correlated
across observations in the cross sectional of the data (Long & Ervin, 2000). The hypothesis was
that;
H1: The data is Homoscedastic.
If the p-value is less than 0.05, the hypothesis is rejected.
Figure Heteroscedasticity Plot
The scatter plots showed that the dots are diffused and therefore concluded that the regression
model does not suffer from heteroscedasticity. In addition the Breusch-Pagan results are
presented in Table .
Table Heteroscedasticity Results
Breusch-Pagan / Cook-Weisberg test for heteroscedasticity
Ho: Constant variance
Variables: fitted values of Employee Performance
chi2(1) = 54.19
Prob > chi2 = 0.328
Source: Field Survey Data (2022)
Results in Table 4.13 show that the p-value is greater than the 5%. Then the hypothesis
was not rejected at a critical p value of 0.05 since the reported Chi2 (1) = 54.19 and p-
value was 0.328>0.05 and thus the data did not suffer from heteroscedasticity.
Normality Test
Test for normality determines if the data is well modeled and normally distributed
(linear). To test the normality of the variables, Shapiro–Wilk test was used as it has the
highest power among all tests for normality. The hypothesis was tested at a critical value
at 0.05, where the rule is that reject H0 if the probability (P) value is less than 0.05 or else
do not reject. The dependent variable should be normally distributed because the study
was analyzed using a multiple regression model where the condition of normality must be
satisfied (Quataroli & Julia, 2012). The hypothesis was that;
H1: The data is normal.
The results for normality are as shown in Table 4.14.
Table Normality Outputs
Shapiro-Wilk
Variables Statistic df Sig.
Employee Performance 0.810 256 0.221
Career Development 0.818 256 0.301
Compensation 0.720 256 0.410
Performance appraisals 0.350 256 0.062
Recruitment 0.639 256 0.512
Source: Field Survey Data (2022)
The results indicated that using the Shapiro-Wilk test of normality, the data is normal
since the p-values are above 0.05 for all the variables and thus we do not reject the
alternative hypothesis (H1). Therefore, the variables on employee performance, career
development, compensation, performance appraisals and recruitment are normal in
distribution and hence subsequent analysis can be carried out.
