Inverse probability weighting for self-selection bias correction in the investigation of social inequality in mortality.

Frequency measures of study participants may not reflect the source population in the presence of self-selection, but the impact on association measures can be limited. IPW may be useful for (self-)selection bias correction, but the returned results can still reflect residual or other biases and random errors.