|
Message/Author |
|
|
Is it possible to specify an item population covariance matrix and the factor covariance matrix (phi) at the same time? For example: Say I have 2 factors each with three items and I specify the item population covariance using the with statement (ie: y1 with y2, etc) can I then specify a phi matrix of the two factors say: 1.00 0.25 1.00 Or do I need to estimate the lambda loadings based on the item covariances first and then just specify a CFA model in the model population command. Thank you. |
|
|
Specifying factors, the WITHs will refer to to the residual covariances among the ys (so you have Psi and a full Theta). So when analyzed as a factor model, it won't be identified. But the data can certainly be generated that way. |
|
Lijin Zhang posted on Sunday, February 23, 2020 - 6:29 am
|
|
|
Hi, When conducting Monte Carlo simulation study about Bayesian approxiamte measurement invariance (BITERATIONS = 100000(10000); 10+ groups), sometimes I got an error for each replication: USE THE FBITERATIONS OPTION TO INCREASE THE NUMBER OF ITERATIONS BY A FACTOR OF AT LEAST TWO TO CHECK CONVERGENCE AND THAT THE PSR VALUE DOES NOT INCREASE. I thought it was because the replications were not convergened. However, I watched each replication carefully and found the psr-value satisfied the criteria. Besides, the simulation study with higher-variance prior converged. Moreover, these errors sometimes disappered when I re-analyzed the simulation study. the simulation results of the simalation with errors and the re-analysis are same. This problem is very common when I doing this simaltion study and I am very interested in the underlying cause. Thank you for your time and consideration. |
|
|
It's a standard message intended as a reminder. If you are certain of convergence, you can ignore it. |
|
Back to top |
|
|