www.cmm.bristol.ac.uk/forum

I have multiple outcome measures from the same patients and have to run multiple simultaneous analyses on the same data. Is there some way I can correct for type 1 error (eg using a different estimator than igls) or some other way to adjust for this? I don't want to do a Bonferroni correction because it is too conservative for my data.

Many thanks

Just to explain my data I have three levels where outcome measures are nested in patients who are further nested in clinicians. The measures are collected repeatedly over 5 time points. I am interested in assessing whether there is a differential change in outcome based on treatment type.

I would imagine if you don't want to do Bonferroni then you could try something like Benjamini & Hockberg's FDR which is less conservative (false discovery rate) and I believe could be used for most models. Don't think there is an estimation procedure related approach like tweaking IGLS here.

Thank you so much.

Would you suggest bootstrapping in this case? I know it does not directly address type 1 error but can provide more accurate estimates and confidence intervals?

Not for fixing the issue of multiple comparisons as bootstrapping will not help with that.