I have a longitudinal dataset with 5 time points (age 18, 26, 32, 38, 45) at which patients had oral examinations. I am looking at the same tooth in each patient in each quadrant at each time point (eg. a mouth has 4 quadrants therefore 4 canines). I have fixed predictors (gender, DMFS index at baseline) as well as well as tooth-level predictors that vary over time (tooth surface codes, presence/absence of tooth, impaction angle) and patient-level predictors that vary over time (smoking status, socioeconomic status). My objective is to model tooth outcome (categorical) at age 45 given these predictors. I am struggling to identify the best approach. I considered survival analysis but it would only allow me to determine whether the tooth is pres/absent at age 45 and I would lose other outcomes like erupted/unerupted, decayed, filled... I also looked at xtgee (couldn't figure out how to do multinomial logistic regression within this function) and I also considered gllamm (appears more complicated than I am capable of managing). Can anyone recommend the most simple approach to fit a multinomial logit model to panel data? gsem?
0 Response to Simple approach to fit a multinomial logit model to panel data?
Post a Comment