Building on 'rimu', a package for inference with multiple-response data
Two main modelling approaches
mrglm: working-independence marginal models (eg Agresti & Liu)mrloglinandmrmultinom: working-multinomial loglinear models (eg Loughin & Scherer)
The first type are glms with a sandwich variance estimator in expanded data. The second type are loglinear or multinomial models in expanded data, and the inference procedures are like Rao-Scott-type tests in survey data
remotes::install_github("tslumley/rata")
library("raataa")