is there any guidance regarding adequate sample size given the no. of intercepts to be estimated? In the Harrell et al. Stat Med paper they consider small (eg n=50) samples in their simulations but estimate “a few intercept values”. And although in the analysis of CD4 counts they estimate 831 intercepts, they have n=4776. I will soon find out because i will estimate CPMs with number intercepts = around 0.5 x n (with n<100). cheers
[i initially posted this Q here: Quality of life surveys and proportional odds models, but maybe more suitable in this thread]