[Question] Plans for the longitudinal binary studies (GEE: glmtoolbox) #25
Description
Dear Authors,
Today I learned about your incredibly useful package. I quickly read the documentation and noticed it supports the single-timepoint case through the glm(). It is not unusual to work with longitudinal studies with binary (and ordinal) endpoint, fit through GEE estimation, reported at more than one timepoint (one is primary, others - secondary) with appropriate "MVT" (based on the multivariate t distribution) adjustment or gatekeeping.
In R, the best available package currently for analysing Gaussian and non-Gaussian Generalized LM, including logistic regression, is the glmtoolbox package (outperforming geepack in stability and functionality), offering also the small-sample Mancl-de Rouen covariance estimator and "talking" with mice, emmeans and several other packages (integration with marginaleffects is pending). Especially the M-dR estimator is priceless, as most of the analyses we did with standard sandwich were rejected as "inappropriate in smaller studies" (where smaller < 300 per arm), and there was no further discussion allowed.
Also, for smaller fractions, the Firth-like penalised GEE logistic regression is an important tool. In R, also this one offers the M-dR estimator via the geesbin package.
Do you, maybe, consider enhancing your awesome package also for longitudinal cases?