|
|
Adjusted chi-square and wlsmv |
|
Message/Author |
|
Jeehoon Kim posted on Friday, February 25, 2011 - 12:24 pm
|
|
|
Dear Professors, I am testing a path model with WLSMV estimator and brr replicate weights. As I didn't get chi-square related fit index with brr replicate weights, I have created an adjusted chi-square statistic by applying design effect with chi-square value from conventional analysis as Stapleton did so (2008). My question is I have used WLSMV estimator to get indirect effects between variables and I wonder how I can get chi-square difference test with this adjusted chi-square value. With diff test from conventional analysis, can I use test result for my test model with replicate weights? Or is there any other way that I can manually test chi-square difference test with adjusted chi-square statistic? I computed CFI, TLI and RMSEA with this adjusted chi-square, and those fit indice are very similar to those from conventional analysis. My fit indice are CFI=0.987, TLI=0.973, and RMSEA=0.042 with adjusted chi-square statistic. Thank you for any advice in advance. Regards, Jeehoon |
|
|
You should use the DIFFTEST on the runs without the replicate weights. Then adjust that chi-square value as in Stapleton (2008) by applying design effect based on the replicate weights. |
|
Jeehoon Kim posted on Saturday, February 26, 2011 - 11:00 am
|
|
|
Thank you for your advice. I would do it. |
|
Back to top |
|
|