All,
I am working on a PK-PD analysis of longitudinal data in a pain study.
I would like to be able make a statement about statistical significance of the
treatment effect at the end of the analysis, thus I would like to have some
control of the type 1 error by having a pre-specified process for how to do
this analysis (although the model may not be 100% pre-specified).
The process I am thinking of is:
1. Develop placebo model for longitudinal data based on placebo data.
2. Fix placebo parameters and fit full data set (including also active
treatment data).
3. Add concentration effect relationship to model and fit the full data
set.
4. Assess whether there was a statistical significant treatment effect
by comparing ofv from 2 and 3.
5. Repeat 2 and 3 but without fixing placebo model.
My questions are:
Is this a sensible approach?
If so should I fix only structural parameters related to the placebo model or
also random effects parameters of the placebo model?
Is step 5 useful? If so what for?
Regards,
Matts Kågedal, AstraZeneca.
--------------------------------------------------------------------------
Confidentiality Notice: This message is private and may contain confidential
and proprietary information. If you have received this message in error, please
notify us and remove it from your system and note that you must not copy,
distribute or take any action in reliance on it. Any unauthorized use or
disclosure of the contents of this message is not permitted and may be unlawful.