One of the major assumptions given for type ordinary least squares regression is the homogeneity in the case of variance of the residuals. In the case of a well-fitted model, if you plot residual values versus fitted values, you should not see any particular pattern. Now, what if the variance given by the residuals is not a constant? In this case, the **residual variance** is called **heteroscedastic**. You can detect the heteroscedasticity in various graphical and non-graphical ways.

The most commonly used way to detect heteroscedasticity is by plotting residuals versus predicted values. In Stata, we can perform this using the `rvfplot`

command. When we leverage the `rvfplot`

command with the option of `yline(0)`

, which is defining the basis of Y equal to *0*, we can see that the data point pattern can get narrower as we move toward the right-hand side. This indicates that heteroscedasticity exists:

**rvfplot, yline(0)**

After running the preceding code, you get the following diagram: