SWOP and CTE require that a correct estimation and association equations are specified.
To determine if the above equations are correctly specified one needs to inspect the residuals (observed-fitted) and determine whether they are roughly equally spaced across the domain of all X-values (although commonly this is across all fitted y-values). This is called random scatter of residuals or heteroscedasticity of residuals.
For example, when fitting a linear model to data that is nonlinear there may not be heteroscedasticity. This absence of heteroscedasticity implores us to find a better fitting equation.