Message/Author |
|
Brian W posted on Monday, April 07, 2008 - 7:43 pm
|
|
|
Hi there, This may be a simple question. I am doing a path analysis and have a categorical indicator variable with 4 levels. From what I have read on here, it looks like I need to create three dummy codes 0/1, just like in regression. Then what? Do I create an ON statement with all dummy codes - i.e., y1 ON dummy1 dummy2 dummy3? How do I interpret the output for these codes? |
|
|
Mplus is a regression-based program. You would create three dummy variables and interpret them as in regular regression. |
|
Xiaoyu Qian posted on Thursday, May 08, 2008 - 10:35 am
|
|
|
What is the maximum number of indicators and factors Mplus can handle using CCFA? I have a test with a lot of items(3 categories) and observations a little over 1000. Thank you for your help. |
|
|
How many is a lot of items? |
|
Xiaoyu Qian posted on Thursday, May 08, 2008 - 10:51 am
|
|
|
The whole test has over 200 items. But the items are ordinal (3 categories) and skewed. So I am thinking about a categorical CFA at first. And probably a second-order CFA later. |
|
|
Your sample size may be too small especially if the items have floor or ceiling effects. You do not want empty cells in the bivariate frequency tables for pairs of items. You might check out these bivariate frequency tables as a first step. |
|
Xiaoyu Qian posted on Thursday, May 08, 2008 - 11:29 am
|
|
|
Thank you. The items are categorized into clusters according to theory. So I did polychoric correlation for one of the clusters using SAS. And I get pretty good correlations (above .53). Not sure if it is equvalent to checking the bivariate frequency tables. So if I want to use a CCFA on this cluster only. Do you think it will work using Mplus? |
|
|
I'm not sure how SAS operates in this situation. You would need to try this is Mplus to see what happens. |
|
Back to top |