Hi,
I'm currently carrying out some multiple imputation.
(I'm using chained-equations but that's not really relevant)
One of my variables is an exclusions criterion for my final analytical
model - respondent exceeds a threshold for probable depression - and
this variable, like many others, suffers from missing data.
Whilst I may feel comfortable that I have other auxiliary data to
impute for this depression measure, I am ultimately looking at
a situation where the final analytical model will be repeated across
imputed datasets of varying size to reflecting the uncertainly in the
true value for this exclusion criterion. It seems to me that this would
then cause problems with parameter SE's when pooling across results as
the n is not fixed.
An additional thought i've just had is that i'm effectively doing
a subgroup analysis here but disregarding the results for those above
the depression threshold. Perhaps I therefore need to incorporate
interactions between depression and my other measures as I would do if I
planned to split by gender.
Anyone have any wisdom to share?
many thanks, Jon
--------------------------------------------------
Dr Jon Heron
Research Fellow
School of Social and Community Medicine
Canynge Hall
39 Whatley Road
Bristol
BS8 2PS
United Kingdom
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|