Skip to content
Prev 181731 / 398503 Next

Hmisc package: deff() command's formula for the design effect

One additional question:

It seems that in practice the design effect is often calculated on the
response variable or on individual predictor variables. I have some OLS
models using observational data that are clustered. Does it make sense to
calculate the design effect on the residuals of one of these models to see
the extent to which there is still clustering left even after including
covariates?
Thomas Lumley wrote: