Hi Bengt/Linda, I have a Q about version 4 (but I understand if you don't want to let any cats out of the bag):
I am running a multilevel path model where I want to estimate the same relationships using the same variables across levels of analysis. However, one of my outcome variables is dichotomous, so model estimation requires numerical integration. Problematically, this means that "Unrestricted x-variables in TWOLEVEL analysis with ALGORITHM=INTEGRATION must be specified as either a WITHIN or BETWEEN variable." So, to estimate the model, I've disaggregate the variance in my x-variables in Excel and then declared the disaggregates as either within or between.
My question is: will version 4 allow numerical integration without declaring variables as either within or between? Also, because I want to estimate indirect effects in my model, will version 4 allow (bootstrapped) indirect effects in TWOLEVEL models?
sorry, but I forgot to also ask if version 4 will have automated two-level chi-square difference testing for models estimated with MLR.
bmuthen posted on Saturday, December 31, 2005 - 11:57 pm
I may be wrong, but I think the x variable issue only comes up when x has a random slope. In that case, the "whole" x (both between and within variation) is what the random slope multiplies. I guess you can also create a between-level part of x as the cluster mean x-bar-subj and use that on between to get different x slopes on different levels. V4 is no different here. Two-level bootstrapping is not going to be in V4.0. We'll see what version 4 will have in terms of nested model testing. It can be safely said, however, that 4.0 will have a long list of new features.
I commonly encounter the x-variable issue without random slopes (does anyone else run into this?). I know the issue can be gotten around by taking x-bar-subj and entering that on the between model part, but then aren't the between parts not "precision weighted" empirical bayes estimates, as they would be if called up to the between-groups level in their original form?
In any event, perhaps I'm doing something incorrectly and am getting the error message as a result. I'll send my input and data to the support address.