Message/Author |
|
shuang posted on Sunday, August 17, 2014 - 8:24 pm
|
|
|
Hi, I'm running a latent interaction model using the following syntax: analysis: type = random; algorithm = integration; model: rel by x1 x2 x3; gx by x4 x5 x6; mod by w1 w2 w3; perf by y1 y2 ye; INT1 | rel xwith mod; INT2 | gx xwith mod; perf on rel gx INT1 INT2; However, the output only shows the loglikehood and information criteria. It did not report Chi-square test, RMSEA etc. As shown below: MODEL FIT INFORMATION Number of Free Parameters 43 Loglikelihood H0 Value -5753.300 H0 Scaling Correction Factor 1.593 for MLR Information Criteria Akaike (AIC) 11592.600 Bayesian (BIC) 11761.112 Sample-Size Adjusted BIC 11624.687 (n* = (n + 2) / 24) Could you advise how I can get the fit index (e.g. CFI, RMSEA)? Thank you! |
|
|
Chi=square and related fit statistics are not available when numerical integration is required. |
|
shuang posted on Monday, August 18, 2014 - 6:53 am
|
|
|
Thanks for the quick response. Could you advise how I could do to get the chi square statistics and other related fit index in this case? |
|
|
Chi-square and related fit statistics are available when means, variances, and covariances are sufficient fit statistics for model estimation. This is not the case with numerical integration. Nested models can be compared using -2 times the loglikelihood difference which is distributed as chi-square. Non-nested models with the same set of dependent variables can be compared using BIC. |
|
shuang posted on Tuesday, August 19, 2014 - 11:58 pm
|
|
|
Thank you very much. Could you pls provide references to test moderation (and associated -2 times the loglikelihood difference)? Very much appreciated! |
|
|
You might look at the Bollen SEM book or David MacKinnon's book. |
|
Back to top |