Homoscedasticity of Residuals Verified through Model Diagnostics
Diagnostic Methods 1
Standardized residual plots can be used to check if a regression analysis was conducted properly. To verify homoscedasticity, one should check if the scatter of the residuals is uniformly distributed overall. Common examples of lack of homoscedasticity include the following two cases.
The variance increases towards the end, a situation that often requires a transformation or the introduction of weights to resolve. Regardless of how easy it is to solve, it’s often the most straightforward and simple solution among the issues discovered during model diagnostics.
Only the middle part has an unbelievably small variance, which could suggest a problem right from the data collection stage. It’s likely that there are other variables that can precisely explain the extreme variance, so it’s recommended to review the dataset once more.
See Also
Hadi. (2006). Regression Analysis by Example(4th Edition): p98. ↩︎