Considerations for the Cog-Cog track

[andkov]

@ampiccinin @smhofer @GracielaMuniz @wibeasley @annierobi

We separated the men and women for the physical analyses due to differences in body size, etc., but it is not standard to separate them in analysis of cognitive data. The only reasons I can think of would be potential differences in educational/occupational background/choices.

Might be worth considering whether it makes a difference in a study like ELSA that has only three cognitive variables – or maybe LASA which has variables I prefer.

The good thing if we don’t separate by sex is that we would have only half as many models. Also, some of the studies have only 3-4 measures, so only 3-6 models. The most would be MAP, with 19 tests (if we did not eliminate “line orientation” and “ideas”). Would that really mean 198 models for MAP alone? EAS, OCTO and SATSA could have as many as 93 each (11 measures). Nuage, with only 1, could not provide any.

[andkov]

by @annierobi
I would suggest we combine males and females as it would mean fewer models and larger samples.

[ampiccinin]

Would it be a waste of time to look at LASA separated and combined as a test case?

[andkov]

The only downside to combining groups is it might disrupt the processing pipeline. However, technical considerations should not influence theoretical ones. Besides, this is a long-overdue expansion. I'll start working in this direction. @beccav8 has been picking up the use of Portland scripts and I meant to adjust them for the use of combined group, so another reason.

Coincidentally, @beccav8 and I have been working on prepping LASA for Portland scripts. I can't guarantee that it will be faster than just doing it manually, but just FYI that we are working in this direction.