Non-constant variance

Now let us look at an example where we might experience non-constant variance.

Assessing model fit we can see from the Residual vs Fitted values plot that the residuals do not appear to have constant variance.

In this case, we could try some transformation of the data. If we initially try to log transform the response variable

This transformation seems to have help rectify some issues with our original model. Please note that this transformation is not guaranteed to be a quick fix.

You may be worried about plot 105 (that can be seen at the bottom of the last two plots). We will address this issue later in this lecture.