Checking the Normality of Residuals through Model Diagnostics
Diagnostics
To determine if regression analysis was performed correctly, you can check using the standardized residual plot.
Normality is better assessed using histograms rather than looking at the scatter of residuals, or by conducting a normality test.
The left side shows a density that decreases towards the top and bottom from the center, whereas the right side is evenly distributed regardless of the direction.
However, cases where the residuals actually follow a known distribution other than normal distribution are almost never found in actual analysis. Issues with normality in analysis are mostly due to outliers.
Normality problems occur when there are considerably more outliers than in the sample or when an absurd number like 6 or 7 appears. Such outliers should not be blindly removed; instead, the analyst must carefully review the data.