Other DDFM options for Fit Mixed (or even Fit Least Squares)

[andrewtkarl]

What inspired this wish list request? 

I am trying to fit a pharmaceutical stability analysis with Fit Mixed (as discussed, e.g., in Statistics for Biotechnology Process Development). However, I cannot replicate the corresponding SAS analysis because the approved methodology uses the SAS default for DDFM (CONTAIN). The current JMP hardcoded variation of Kenward Roger (which appears to be accurate and similar to analogous PROC MIXED output) gives me volatile DF estimates that render the current implementation in JMP unusable. I understand the benefits of KR, but this is a set procedure (sort of like the corresponding fixed effect stability analysis in the Degradation platform) that I need to match.

What is the improvement you would like to see? 

Add at least one of the other DDFM options similar to PROC MIXED for Fit Mixed and/or Fit Least Squares.

Why is this idea important? 

This is a common application for pharmaceutical customers and they may need to use SAS instead.

[SamGardner]

@andrewtkarl thanks for bringing this to our attention.  We will look into this and let you know what we decide to do.  

[SamGardner]

@andrewtkarl can you point us to the reference you mentioned, "Statistics for Biotechnology Process", and any other relevant Pharma guidance that we can use as we investigate this?  

[andrewtkarl]

Thanks, @SamGardner ! 

• Section 7.4.2 of Statistics for Biotechnology Process Development, CRC Press, Edited by Todd Coffey and Harry Yang (chapter written by Laura Pack)
• Section 8.4.4 of Statistical Applications for Chemistry, Manufacturing, and Controls (CMC) in the Pharmaceutical Industry, Springer, by Richard K. Burdick et al.
• Section 10.5 of SAS for Mixed Models: Introduction and Basic Applications, SAS Press, 2018, Walter Stroup et al.
• ICH Q1E is the guiding principal for how the Degradation > Stability platform works. This guidance document also mentions that it is allowable to treat the batches as random effects. So these sources are various attempts to do so.

[andrewtkarl]

.