Student 09 posted on Saturday, October 03, 2009 - 9:18 am
in a twolevel model with a binary dependent variable y1 and a continuous predictor x1 - which I would like to decompose in its within- and between-group regression coeffcients - I received the following error message:
Unrestricted x-variables in TWOLEVEL analysis with ALGORITHM=INTEGRATION be specified as either a WITHIN or BETWEEN variable. The following variable cannot exist on both levels: x_1
Are there any alternatives to circumvent this restriction? I don't like to specify x_1 either/or as within/between given that its between-variance is large and I am looking for contextual effects?
It sounds like you have specified a random slope, in which case the x variable is not decomposed. See UG ex 9.2. This is exactly like in regular multilevel modeling. It does not mean that x does not have between-level variation. If you don't have a random slope, send your input, output, data and license number to firstname.lastname@example.org.
You can enter as a between-level version of x its cluster mean.
There are also possibilities for a latent variable decomposition.