Yung Chung posted on Thursday, June 21, 2007 - 12:28 am
Hi, good evening:
I estimated a model where one of the outcome variables is a binary variable (four-year college enrollment or not: enroll4) and everything else was treated as continuous. Using the command "categorical are enroll4", the weighted least square with mean and variance adjustment was therefore invoked.
Am I correct in my assumption that Mplus analysed tetrachoric (?) correlation matrix opposed to covariance matrix?
When you have a combination of continuous and categorical variables, the matrix that is analyzed has different entries depending on the scale of the variables. For two continuous variables, the off-diagonal entry is a covariance and the diagonal entries are variances. For two binary variables, the off-diagonal entry is a tetrachoric correlation and there is no diagonal entry. For a continuous and a categorical variable, the off-diagonal entry is a biserial correlation and there is no diagonal entry.