Jong Min posted on Friday, April 14, 2006 - 7:43 pm
Can I get some help with the following data analysis question?
I have 8 indicators that are measured in a mixture of nominal and ordinal levels. Those questions are all related to assessing subjective "health status."
Q1. with 5 ordinal categories Q2-Q3. with 3 ordinal categories Q4-Q5: yes/no Q6: with 5 ordinal categories Q7-Q8: yes/no
What is the best way of reducing these different mixed type of indicators? Traditional EFA would not work in principle due to their levels of measurement. I wonder if using EFA/CFA with categorial variables in Mplus would be appropriate. Can I calculate Factor Scores in Mplus?
Eventually, the reduced variable will be used in the context of regression, with replicative weights applied in Stata program.
Any suggestions or comments would be greatly appreciated.
Binary and ordered categorical variables can be considered in EFA. I'm unclear about what you mean by "levels of measurement".
Jong Min posted on Friday, April 14, 2006 - 8:11 pm
Thanks for your quick reply. Sorry for confusion. By "level of measurement," I meant that traditional EFA is for either interval or ratio levels of measurement, but variables that I have were measured as nominal or ordinal level of measurements. I was hoping to get some suggestions on how to best reduce a mixed group of nominal and ordinal variables into one variable...
I am doing a CFA with both binary and ordered categorical variables using Mplus 2.13. In the output, I did not see the threshold information for the categorical variables. Was this information available in this version of Mplus?
Below are the barebones of my syntax for reference.
variable: ; usevar are ; categorical are ; analysis: type = general; MODEL: factor1 BY ; factor2 BY ; factor1 WITH factor2; OUTPUT: sampstat standardized ;