Categorical variables in lavaan, please clarify

[Elias Carvalho]

As a newbie in lavaan I got confused with categorical data.Somebody would clarify?

If I have:
Exogenous variables

• 1 binary (Gender: 0 MALE 1 FEMALE)
• 1 ordinal (BMI: 1 NORMAL, 2 OBESE, 3 OVERWEIGHT)
• 1 nominal (RACE: 1 WHITE, 2 BLACK, 3 OTHERS)

Question: In this case I recode GENDER and RACE as dummy and tranform it in numeric and for BMI only transform in numeric?

Endogenous variables:

• 1 Binary variable (HAS DIABETES: 0 NO 1 YES)
• 1 ORDINAL variable (PAIN LEVEL: 1 - NO PAIN 2 MODERATE 3 SEVERE)

Question: Keep all as categorical, but tranform it as ordered factor and use

pa.model <- SEM(myModel, data = myData,ordered=c("DIA","PAIN"))? 

And nominal variables as endogenous? How to work with them?

[Terrence Jorgensen]

I recode GENDER and RACE as dummy and tranform it in numeric 

Yes

 

and for BMI only transform in numeric?

Only if you can assume the effect is linear (same change between 1 and 2 as between 2 and 3).  I would recommend using the middle category as a reference and make a dummy for the upper/lower categories.  Label their slopes if you want to test linearity (i.e., see if their absolute magnitudes are equal).

Endogenous variables: Keep all as categorical, but tranform it as ordered factor and use

pa.model <- SEM(myModel, data = myData,ordered=c("DIA","PAIN"))? 

Yes

And nominal variables as endogenous? How to work with them?

You don't, not in lavaan.  You could only recode into a dummy to predict being in category 1 vs. any other category, but you would not be able to go beyond binary regression for a nominal outcome.

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen