Looking at the variation between untreated individuals in affected vs unaffected (SMS-unt vs Con-unt)

For filtering, I am doing CPM with CPM < 5 in at least 3 samples (should I use a different method?)

What options would most accurately depict the variation in expression

Fit Type: Parametric vs Local vs Mean

Beta Prior: True vs False

Test Type: LRT vs Wald

Do you have any suggestion on what criteria would provide me with the most accurate results - and why

Am filtering the results to a padj < 0.05 and a fold change of 1, then doing enrichment on that gene set

Clone Disease Treatment Age
AT.1    A1  SMS   VPA       Y
AT.2    A2  SMS   VPA       Y
AT.3    A3  SMS   VPA       Y
AU.1    A1  SMS   Unt       Y
AU.2    A2  SMS   Unt       Y
AU.3    A3  SMS   Unt       Y
OT.1    O1  Con   VPA       Y
OT.2    O2  Con   VPA       O
OT.3    O3  Con   VPA       O
OU.1    O1  Con   Unt       Y
OU.2    O2  Con   Unt       O
OU.3    O3  Con   Unt       O