I was interested to run a path analysis comparing males versus females. I am also using complex weights and I wasn't sure how to write the syntax. I have a PSU, Stratum and the sampling weight...is the syntax below correct? I wasn't sure what to consider PSU and stratum.
VARIABLE: NAMES ARE v1 v2 v3 v4 v5; USEVARIABLES ARE v1 v2 v3 v4 v5; WEIGHT IS v2; (is this the sampling weight from my data set?) STRATIFICATION IS v4 (is this the PSU?) CLUSTER IS v5 (is this the stratum?)
ANALYSIS: TYPE IS COMPLEX; ESTIMATOR IS MLR; ITERATIONS = 1000; CONVERGENCE = 0.00005;
I know the model part but just wanted to make sure I had the weighting part correct. Thank you.
You need to continue the record on a second line so you do not exceed 90 characters. If you cannot see how to do this and continue to have problems, send your input, data, output, and license number to firstname.lastname@example.org.
I am using MLR and SB factor to compare models. I read somewhere that the fit indices also need to be adjusted as well. I looked on your website but I only found the calculation for the adjusted chi-square. Do you know if there is and where I could find the calculations for the other fit indices (CFI, RMSEA, etc)? Thank you.
I am conducting a path analysis with multiple group comparison. I know I have to report the unstandardized coefficients as opposed to the standardized ones but could you explain to me why that is? Also could you maybe provide me with a reference or two where I can learn more about this issue? Thank you.
The basic issue with standardized coefficients in multiple group analysis is that for raw coefficients that are constrained to be equal across groups, the standardized coefficients are not equal because the standardization uses group-specific information. For more information, Google Sander Greenland who has written on this topic.
Hi, I am running a similar model to those described above – a path model with stratification, cluster and weight variables. The main difference is that I have a binary dependent variable. I can run the model for the whole sample with MLR (with montecarlo integration), but have had to use WLSMV when I use the GROUPING command to assess gender differences. Unfortunately some key coefficients differ between the MLR and WLSMV models (I suspect due to the difference between FIML and pairwise present methods for each estimator). Is there any way I can run my model using a FIML based estimator and test gender differences? Many thanks in advance!
Variable: WEIGHT is Dweight; CLUSTER is SampPSU; STRATIFICATION is SampStra; Names are W7_wt ks4 zks4 Dweight SampPSU SampStra W3finwt female uni1dk uni2dk uni3dk uni1 uni2 uni3 eff1 eff2 eff3 atuni zeff1 zeff2 zeff3 zuni1 zuni2 zuni3 zks3 znssec3 zhiqual zincome zks2 russ prest zprest; Missing are all (-9999) ; CATEGORICAL=atuni; USEVARIABLES ARE atuni zuni1 zuni2 zuni3 zeff1 zeff2 zeff3 znssec3 zhiqual zincome zks2 zks4;
Analysis: TYPE IS COMPLEX; ESTIMATOR IS MLR; INTEGRATION=MONTECARLO;
Thanks Bengt, When I run the model with type=mixture and knownclass the models won't run - first I get a warning message saying it does not support INTEGRATION=MONTECARLO and when I try ALGORITHM=INTEGRATION as the warning messages suggest I get the following message: *** FATAL ERROR THIS MODEL CAN BE DONE ONLY WITH MONTECARLO INTEGRATION.
I can run the models if I do not specify my dependent variable as categorical - is this a viable option for multiple group testing?