|
|
 |
| AVEPP's for manually generated cluste... |
 
|
| Message/Author |
|
|
|
|
Hi!
I have done a RMLCA based on five years where five variables have been used, i.e. one dichotomized, two nominal and two ordinal. Data are in long format and looking at the output file from MPlus it could be observed that some individuals change their cluster belonging over time. My coauthors are interested in manually creating new clusters in addition to some of those found from the cluster analysis. Is this a problem later on when our interest is to do a Cox regression and as an explanatory variable using both these types of clusters? Specifically, could average posterior probabilities be calculated for those clusters not generated from the analysis and be used in a 3-step?
Best wishes!
Magnus |
|
|
|
|
| I think this is a good idea. If you have say 2 classes, and 2^5=32 patterns, you can combine the "unstable" patterns into class 3 and recompute all the steps in the 3-stage methodology. |
|
|
| Back to top |
|
|
