Uji Asumsi Klasik

Classical Assumption Tests

Classical assumption testing involves checking whether research data meet the conditions required by the statistical method being used. For regression, these checks help assess whether model estimates are reliable, unbiased, and consistent. If assumptions are not met, data transformations may be considered.

Assumptions discussed

• Linearity: The relationship represented by the regression model should be linear.
• Normality: The normality test is applied to the model’s residuals, not separately to each original variable. The assumption is met when residuals are normally distributed or approximately so.
• Multicollinearity: This checks whether independent variables are correlated with one another. The lecturer mentions the Variance Inflation Factor (VIF) and states that a VIF above 10 indicates a multicollinearity problem.
• Heteroscedasticity: This checks whether residual variance differs across observations. A model without heteroscedasticity is described as homoscedastic.
• Autocorrelation: This checks whether the error in one period is correlated with the error in the previous period. The Durbin–Watson (DW) statistic is mentioned as a measure used for this test.

Normality testing with Kolmogorov–Smirnov in SPSS

The lecture demonstrates a normality test using illustrative data, not real research results. The example includes a dependent variable, Y, and independent variables such as X1, X2, and X3. The demonstration uses a simple regression with Y and X1.

1. Enter the variables and data in SPSS.
2. Choose Analyze → Regression → Linear.
3. Assign Y as the dependent variable and an independent variable—in this example, X1—as the predictor.
4. In the regression options, save the model’s residuals.
5. Choose Analyze → Nonparametric Tests → Legacy Dialogs → 1-Sample Kolmogorov–Smirnov.
6. Select the saved residual variable and choose the normal distribution for the test.
7. Compare the reported significance value (p) with the chosen alpha level, commonly 0.05:
   • If p is greater than alpha, the residuals meet the normality assumption.
   • If p is less than or equal to alpha, the normality assumption is not supported.

The example reports a significance value of 0.874, which is greater than 0.05. The lecturer therefore concludes that the residuals are normally distributed and the normality assumption is met. Students are encouraged to practise testing for multicollinearity, heteroscedasticity, and autocorrelation using SPSS.

Speakers and sources

• Devi Nurtiyasari is the lecturer and sole audible speaker.
• SPSS is demonstrated as the statistical software tool; it is not a speaker.