Structural Equation Modeling (SEM) Discussion > Combination of ordinal and binary variables
Hi Joseph! I think the AMOS development video tutorials page has a whole video on doing SEM with binary and categorical data! Check it out at the URL below:
http://www.amosdevelopment.com/video/index.htm
I hope that helps!
February 11, 2013 |
Jeremy Taylor
Jeremy Taylor
I have a data set with quit a number of variable and hoping to see if I can get more out of it by using SEM. I have a survey of with all likert scales which range from 1-7 and are about 6 concepts which have 3 - 6 items each and there is one decision which is dichotomous/binary. So basically I have 6 latent variables with 3-6 observed variables each and want to see if they predict what choice is made. Now I keep reading that using different kind of data, ordinal and continous, requires different approaches , but I can't find what to do with binary data.
I did read that one way of using binary variables is using the ADF estimation, however for a sample size of more than 2000 is advised, while I have a sample size a little below of 300. Additionally I've seen papers on the transformation of binary variables, but noticed instruction books on SEM don't mention anything about this.
Can I analyze this data using SEM, because of the difference in data type and low sample size? if yes, how should I do it? Also should I model a path from the latent variables to observed binary variable directly or add a latent variable between it (with only the binary variable as observed variable) .
Currently I'm using LISREL, altough I also have acces to R and AMOS.