www.cmm.bristol.ac.uk/forum

I would like to estimate a multilevel discrete time survival analysis with dummy variables for each of the time points. I would like to specify the model by including all TIME dummy variables rather than omitting one of them and including an intercept. Reardon, Brennan, & Buka (2002) specify this model in equation 4 on p. 307 and note that "because there is no omitted age dummy variable, Model 4 includes no y00 intercept term in the level-two equation for alpha j0." However, in MLwiN I seem to be able to only set up this model with an omitted dummy variable. Ultimately, I would like to estimate a model with a cross-level interaction between TIME and a higher level predictor, as in equation 6 in Reardon et al. But again, I would like to have all TIME dummy variables in the model. How do I do that in MLwiN? An additional complication is that I have a 4-level model (i=time, j=individuals, k=communities, l=community pairs).

My model looks like in the attached file. TIME_0 is the reference category. ECTYPE is a level 3 (community) predictor.

Thanks!

Sabrina Oesterle