5
$\begingroup$

I have read other posts on conducting Factor analysis (FA) with dichotomous variables and although it appears clear that FA done in the default way is not appropriate, I am still unclear about a few things.

I have 7 variables that are yes/no responses. They are assessing peoples functioning (e.g., working yes/no, involvement with justice system yes/no etc). I want to combine these responses to have two variables that represent domains of functioning (e.g., occupational and social). I suspect three of them will be occupational and 4 of them social. I would like to do something analogous to a FA to confirm this.

I only have access to SPSS (and am only familiar with this package) so am looking for advice on how to do this with SPSS.

Do I do a FA with tetrachoric correlations instead of Pearsons (how is this done?)? Or can I do Latent Trait Analysis in SPSS?

I am specifically looking for references with step-by-step instructions.

$\endgroup$
1
$\begingroup$

I would probably recommend the LCA route. Penn State's Methodology Center has a standalone program that I believe is free you can use to analyze the data (WinLTA). I have used their LCA/LTA analysis package for SAS and have found it to be quite good.

$\endgroup$
1
$\begingroup$

If you install the (free) R Essentials for SPSS Statistics (available from the SPSS Community site at www.ibm.com/developerworks/spssdevcentral via the Downloads for SPSS Statistics link, you get the STATS LATENT CLASS command and associated dialog box for it.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.