Message/Author |
|
Lan Huang posted on Wednesday, October 21, 2009 - 8:17 pm
|
|
|
Hi, I'm running a CFA and my observed variables include continuous variables and dichotomous variables. And there are values missing by design in both kinds of variables. I'm try to use the PATTERN option but I got this error message:"Analysis with categorical variables is not available with PATTERN, COHORT, COPATTERN features." My questions are: 1. What's the function of PATTERN exactly? Just rearrange the data so cases with the same missing patterns will be organized together? Or something else? 2. If the PATTERN option can't be used with categorical data? Do you have any suggestions to deal with this kind of missing by design? Any idea will be appreciated! Thank you! Lan |
|
Lan Huang posted on Thursday, October 22, 2009 - 6:52 am
|
|
|
A correction: And there are values missing by design in only the dichotomous variables. Thank you! |
|
|
You do not need to use the PATTERN option. Just use the default which is TYPE=MISSING and maximum likelihood estimation if you want FIML. The default in your case would be weighted least squares estimation. |
|
Lan Huang posted on Friday, October 23, 2009 - 12:42 pm
|
|
|
Thank you, Linda. I just want to clarify my question. In our design, data are missing by design as described on page 455 of the manual. There are two forms of a test with 4 linking items. Consequently, there are some covariances for variables X and Y that cannot be computed because items X and Y do not appear on the same form. When I said below that there are missing data, I also meant that some of the covariances are missing. Will the approach that you describe above still work? Complicating matters is that some of our variables are continuous and some are categorical. The 4 items common to both forms are categorical (dichotomous). Thank you so much. Happy Friday! Lan |
|
|
Coverage for some covariances can be zero. These parameters cannot of course be estimated. This can be used with a combination of categorical and continuous outcomes. |
|
Back to top |