Hello!
I seeking advice on how to calculate goodness of fit (GOF) statistics for a SEM model that has complex survey data (i.e. includes sampling weights) AND uses full information maximum likelihood to address missing values? It appears that traditional measures of GOF, like the CFI, TFI and RMSEA cannot be calculated in Stata when sampling weights are included. The output says "Note: model was fit with vce(robust); only stats(residuals) valid."
This still leaves the Standardized Root Mean Square Residual (SRMSR) though, which is a commonly recognized measure of GOF--however, when used with FIML approaches, it appears SRMSR cannot be calculated in Stata and the output says: "SRMR is not reported because of missing values."
Put them together and I cannot get GOF statistics. I have Stata 15.
Any advice on how to generate a GOF statistics reviewers will recognize for a model that uses sampling weights and FIML?
Thank you,
Matt
I seeking advice on how to calculate goodness of fit (GOF) statistics for a SEM model that has complex survey data (i.e. includes sampling weights) AND uses full information maximum likelihood to address missing values? It appears that traditional measures of GOF, like the CFI, TFI and RMSEA cannot be calculated in Stata when sampling weights are included. The output says "Note: model was fit with vce(robust); only stats(residuals) valid."
This still leaves the Standardized Root Mean Square Residual (SRMSR) though, which is a commonly recognized measure of GOF--however, when used with FIML approaches, it appears SRMSR cannot be calculated in Stata and the output says: "SRMR is not reported because of missing values."
Put them together and I cannot get GOF statistics. I have Stata 15.
Any advice on how to generate a GOF statistics reviewers will recognize for a model that uses sampling weights and FIML?
Thank you,
Matt
