Thanks

Hello Ceren,

The variables for factor analysis are presumed to be continuous, not discrete. That said, many software packages will not balk at the submission of discrete variables (hence, Young-Rock's question about which software you were using).

However, for discrete variables, polychoric (of which tetrachoric is a special case) correlations are recommended, since they estimate what the underlying continuous form of the variables might manifest as correlations. Here is a link to a free factor analysis package that can compute and use polychoric correlations: http://psico.fcep.urv.es/utilitats/factor/

Good luck with your work.

Thanks for your help. I use SPSS. Can not I use confirmatory factor analysis using AMOS to test validity of my construct?

The short answer is, yes. The longer answer is, neither SPSS base nor the AMOS add-on compute polychoric correlations, so you'll be working with good old Pearson correlations (which tend to underestimate tetrachoric values).

Thanks

You could also look into the program Mplus, which have all sorts of capabilities, including the one you are looking for, and which is very easy to use. It is, however, not a free program.

Check the web site www.statmodel.com.

Hello,

The best way is to use R and the Lavaan package (even if Mplus do the same). Factors extraction won't be based on the Bravais-Pearson correlations matrix but on the polychoric correlations matrix. It is very easy to calculate. See the Lavaan tutorial: https://dornsife.usc.edu/assets/sites/210/docs/GC3/lavaan_tutorial.pdf (chapter 10: Using categorical variables).

With Mplus, you have to declare your categorical variables ( CATEGORICAL ARE item1 item2....) and use the WLSMV estimator. It will be exactly the same job than Lavaan. The graphics are better with Mplus. Short Mplus tutorial: https://stat.utexas.edu/images/SSC/Site/documents/MPlus_Tutorial.pdf (chapter 5.3. Confirmatory Factor Analysis with Categorical Outcomes).

Best and good luck

CH