Fred Miao posted on Friday, September 14, 2007 - 6:08 am
I have a categorical variable "career stage" and I am wondering if I can include it as an independent variable in my model. Career stage is measured with four descriptions (stage 1=exploration; stage 2=establishment; stage 3=maintenance; stage 4=disengagement). While these four measures are descriptive in nature, in theory they are "continuous" in the sense that one goes through stages 1-2-3-4.
Could I then treat career stage like a 4-point Likert scale and put it in my model as an IV ?
The scale of an independent variable does not matter in model estimation. Independent variables can be binary or continuous. In both cases, they are treated as continuous in model estimation. You can treat the variable as continuous or create a set of dummy variables to represent the categories.
I am working on a research project which I should use SEM method to confirm my model. I performed the analysis in two different ways: First I performed an explanatory factor analysis on the results of questionnaires and reduced 74 indicators (which were measured in Likert scale) to 14 factors. Then I used factor scores to develop a SEM model with continuous variables (14 indicators and 6 latent variables). The overall fit of the model was OK according to Chi-square, CFI, and other criteria (P-V=0.25) but 3 of the relations were not significant, while they must be according to the logic and literature in the field of research! Second, I used 74 indicators as categorical variables and using WLSMV of Mplus software performed the analysis. In this case P-V= 0.0000 but non-significant relations of the first model are now significant. As I am not familiar enough with this kind of SEM, I will be so happy if you guide me in this case. Should we check the adequacy of the second model like SEM models with continuous variables? How about the criteria of model adequacy checking in the SEM with categorical indicators? Can you introduce me an article? Why usage of SEM with categorical indicators is rare in social sciences? Isn't it strong enough or it is new?
I think your question is how to assess fit when dependent variables in SEM are categorical. You do this the same way as with continuous dependent variables. See the Yu dissertation on our website for further information about this topic.
dear dr. Muthen i want to apply 6 types of models on SEM so i need your help to write these commands in mplus. 1- independent continuous variables. 2- independent ordered categorical variables. 3- independent dichotomous variables. regards
Dear dr. Linda I am sorry I mean observed variables not not independent because when I used mplus I saw this term "weight variable" and "grouping variable"for one variable only and I can select categorical variable to all variables not only one like grouping and weight variables then how can I analyze more than one variables and how can I select continues variables. Grouping var. Means dichotomous? And categorical var. Means ordered? Then what about continues how can I chose it. Regards
I read that I can create separate dummy variables in MPLUS using the define command. I just want to make sure I am doing it correctly.
An example of one of my variables as currently listed is: Race 0: white 1: Asian 2: black 3: other
Is this the correct format for my code? DEFINE: white = 0 if (race eq 0) then white = 1 asian = 0 if (race eq 1) then asian = 1 black = 0 if (race eq 2) then black = 1 other = 0 if (race eq 3) then other = 1
I don't think this is right as I do not have k-1 variables listed above.
Any help you can offer would be greatly appreciated