|
|
Dcat/dcon vs. r3step for LCGA |
|
Message/Author |
|
Julia Tang posted on Wednesday, September 03, 2014 - 3:17 pm
|
|
|
I'm running at LCGA and was wondering, if I'm really just interested in looking at covariates descriptively, is it acceptable to look at dcat/dcon output, even if the variables are measured earlier in time? Specifically, if I'm trying to understand significant differences in classes on a series of demographics, but I am not hypothesis testing (i.e. just trying to identify the proportion of males vs. females within classes), can dcat/dcon be used instead of r3step? |
|
|
Yes, this is acceptable. |
|
|
Does the DCAT method change class membership when estimating distal outcomes? Also, it is correct that the CATEGORICAL command is not needed for specifying the measurement scale of the auxiliary vars? |
|
|
Q1: It can but typically not much. Q2: No. |
|
|
Thank you for your quick reply. May I also ask whether it is appropriate to apply Bonferroni correction to the DCAT results when there are >2 classes? |
|
Back to top |
|
|