|
|
Added covariance in two-level model |
|
Message/Author |
|
|
Hello, I'm running a number of multilevel models, and I wanted to clarify one portion of the output. The thing that's curious to me is that in the "between level" output (below), a covariance between lang2 and lang3 is added (LANG3 WITH LANG2). This is not something I specified in the input. I haven't been able to identify the reason for this. Do you know what is happening here? Find input/output attached |
|
|
Many parameters are estimated as the default. If you don't want the parameter, you can fix it at zero, for example, LANG3 WITH LANG2@0; |
|
|
Hello, I have a follow up question about this. I am trying to figure out whether the lang3 with lang2@0 command is correlating the observed values or the error variance at zero. I read in another post (http://www.statmodel2.com/discussion/messages/9/731.html?1190114651) that it depends on whether the variables are exogenous or endogenous. I'm not sure of the answer in this case given that lang2 predicts lang3 in the "within" level of the model, but lang2 is also predicted by a number of other variables. I believe this means they are both endogenous. Is this correct? If so, does this mean the correlation is between the observed values, not the residual variances? Thanks. |
|
|
Yes, both of them are endogenous and therefore WITH refers to their residual covariance. |
|
Back to top |
|
|