Diagnostic Test

Diagnostic test
Before conducting regression analysis, the researcher performed diagnostic tests to

assess normality, heteroscedasticity, multicollinearity, autocorrelation, and
linearity.
Normality Test
The test statistic in a regression model should follow a probability distribution that
is easy to handle, such as the normal distribution. Normality can be assessed by
examining the residuals to determine if they follow a normal distribution (Jones,
2020).
Normal Q-Q plot Normal Data

Normal Q-Q plot Right-Skewed Data
Normal Q-Q plot Left-Skewed Data
Normal Q-Q plot Heavy-Tailed Data
Normal Q-Q plot of Light-Tailed Data
Test for Normality using Shaphiro-wilk Test
The Shapiro-Wilk test for normality generates a p-value that determines whether
the data follows a normal distribution. All predictors and dependent variables are
tested using the Shapiro-Wilk test. If the p-value is greater than 0.05, the data is
considered to be normally distributed (Shapiro & Wilk, 1965). The table
demonstrates that all variables had p-values exceeding 0.05, indicating normal
distribution for each variable. Consequently, further statistical analyses can be
conducted on the data.
Shaphiro-wilk Test
Variables stats df Sig.
Performance of .984 160 .056

the Firm
Technology .997 160 .988

integration
Customer .986 160 .108

integration
Product .989 160 .246

integration
Process .994 160 .755

integration
a. Lilliefors Significance Correction

Multicollinearity
Multicollinearity was assessed by examining the Variance Inflation Factor (VIF)

and tolerance (1/VIF). According to Schwarz, Schwarz, and Black (2014), if the
VIF value exceeds 3, the independent variables are considered collinear. The VIF
values in the table below range from 1.378 to 1.956, indicating that there is no
issue with multicollinearity in the data.
Test of multicollinearity
Collinearity Statistics
Model Tolerance VIF
Technology integration 0.511 1.956
Customer integration 0.717 1.394
Product integration 0.600 1.665
Process integration 0.726 1.378
a. Dependent Variable: Performance of the Firm
Heteroscadacity
Heteroscedasticity occurs when the variance of errors changes across data, leading
to an inefficient yet unbiased OLS estimator (Long & Ervin, 2000). To assess
heteroscedasticity, the researchers utilized the Breusch-Pagan and Koenker tests,
which test the null hypothesis that the variances of the error terms are constant. If
the significant result is less than 5%, the null hypothesis is rejected. The findings
of the Breusch-Pagan and Koenker tests are presented in the table below, showing
that heteroscedasticity was not an issue, as all significant values were greater than
5%.
Breusch-Pagan and Koenker test statistics and sig-values
Tests LM Sig
BP 1.089 .580
Koenker 1.211 .546
Autocorrelation
The autocorrelation test is used to determine if there is a linear relationship

between the errors in a series of observations sorted by time (time series). The
presence of autocorrelation in error terms significantly reduces the model’s
accuracy. This test is essential for time series data to ensure that there is no
autocorrelation. One of the assumptions of linear regression is that the residuals are
independent. To test if this assumption is met, it is necessary to examine a residual
time series plot, which is a plot of residuals versus time (Mbati & Osoro, 2020).
Ideally, most of the residual autocorrelations should fall within the 95% confidence
bands around zero, approximately ±2 divided by the square root of n, where n is
the sample size. Autocorrelation can be formally tested using the Durbin-Watson
test. The Durbin-Watson (DW) statistic ranges from 0 to 4. A DW value of 2
implies no autocorrelation, a value between 0 and 2 indicates positive
autocorrelation, and a value between 2 and 4 indicates negative autocorrelation.
Additionally, one can examine the residual versus time plot to identify any
seasonal or correlated patterns in the residual values (Mkonu & Gichana, 2019).
The autocorrelation was tested using the Durbin-Watson test. The results are
presented in the table below.
Durbin-Watson results
Adjusted RStd. Error of the

Model R R Square Square Estimate Durbin-Watson
1 .802a .650 .555 .60554 1.742
a. Predictors: (Constant), Technology Integration, Customer Integration, Product

Integration, Process Integration
b. Dependent Variable: Firm Performance
The Durbin-Watson statistic always assumes a value between 0 and 4. A DW value

of 2 indicates no autocorrelation, values below 2 indicate positive autocorrelation,
and values above 2 indicate negative autocorrelation (Corporate Finance Institute,
2022). The Durbin-Watson statistic for this study is 1.742. Since this value is close
to 2, it suggests that the errors are independent, indicating no autocorrelation.
Linearity
The linearity test assesses whether a significant linear relationship exists between
two or more variables. In linear regression, the assumption is that there is a linear
relationship between the independent variable x and the dependent variable y. To
verify this assumption, a scatter plot of x versus y is created. This visualization
helps determine if the data points align along a straight line, indicating a linear
relationship between the variables and confirming that the assumption is met (Kusi
et al., 2014).
The results from these tests are used to guide decisions about selecting the
appropriate regression model. If the scatter plot of x and y suggests a lack of a
linear relationship, possible adjustments include applying a nonlinear
transformation to the variables. Common transformations involve taking the
logarithm, square root, or reciprocal of the variables. Alternatively, adding another
independent variable to the model might be beneficial. For instance, if the scatter
plot shows a parabolic pattern, including x2 as an additional independent variable
could improve the model (Kamar, 2019).
Linearity testing can be performed using the significance values for linearity and
deviation from linearity provided in the ANOVA table (Ngechu, 2020). The Sig.
linearity value indicates how well the independent variable follows a straight line.
If Sig. linearity is less than the significance level (α), this suggests that linear
regression is appropriate for explaining the relationship between the variables.
Conversely, Sig. deviation from linearity assesses whether the data can be
considered linear. If Sig. deviation from linearity is greater than the significance
level (α), linear regression remains suitable for analyzing the variables (Smith,
2019). Testing for linearity determines if there is a mathematical relationship of the
form y=cx, where 𝑐c is a constant. This is important because many statistical
methods assume linearity of the data (Cooper & Schindler, 2003).
The following figure indicates a linear trend
The following graph suggests a nonlinear trend

Diagnostic Test

Uploaded by

Copyright:

Available Formats

Diagnostic Test

Uploaded by

Document Information

Original Title

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

Diagnostic Test

Uploaded by

Copyright:

Available Formats

Diagnostic test

Before conducting regression analysis, the researcher performed diagnostic tests to

Normal Q-Q plot Normal Data

Variables stats df Sig.

Performance of .984 160 .056

Technology .997 160 .988

Customer .986 160 .108

Product .989 160 .246

Process .994 160 .755

a. Lilliefors Significance Correction

Multicollinearity was assessed by examining the Variance Inflation Factor (VIF)

Model Tolerance VIF

Technology integration 0.511 1.956

Customer integration 0.717 1.394

Product integration 0.600 1.665

Process integration 0.726 1.378

a. Dependent Variable: Performance of the Firm

Breusch-Pagan and Koenker test statistics and sig-values

The autocorrelation test is used to determine if there is a linear relationship

Adjusted RStd. Error of the

1 .802a .650 .555 .60554 1.742

a. Predictors: (Constant), Technology Integration, Customer Integration, Product

b. Dependent Variable: Firm Performance

The Durbin-Watson statistic always assumes a value between 0 and 4. A DW value

The following figure indicates a linear trend

The following graph suggests a nonlinear trend

You might also like