Dear allstaters
I have a query that i wondered if you could help me with. I am working on
a paper that has come back from the reviewers with major revisions to be
made. one of the points is that the sampling fractions in the study is
greater than 5% and the reviewer wants me to take account fo this in my
analyses. To be honest it makes very little difference (as it is not that
much greater than 5%). I am running a multilevel logistic regression and
have adjusted the errors of the parameter coefficients for the fixed part
of the model. But i am not sure what i am supposed to do about the random
part of the model.
I am looking at the impact of adding groups of variables to the
unexplained residual variation for the different levels of the model and
calculating a ICC for each level. Would the sampling fraction adjustment
alter the residual variance and/or its SE? i imagine it might. Would it
alter relative patterns of variation as well as the size, eg ICC? would it
affect a statement such as "the residual variation at the school level was
halved by the the addition of gender to the model"? ie relative impact of
variables?
Is there any standard way of making the adjustment?
Any help would be greatly appreciated!!
katy Levin
University of Edinburgh
|