Need to standardize variables before SEM for CFA?

[Victoria Lorvig]

Hello All,

I am using structural equation modelling( SEM) for confirmatory factor analysis (CFA). My objective is to make indicators for personality traits to be used in a binary logistic regression for if someone will be willing to change their behaviour or not. The final regression will have the latent variables along with demographics as the independent variables.

The observed variables being used to form latent variables are either likert scale questions (on scales 1 to 5 or 1 to 10) or binary (yes or no). For plain old factor analysis, I was standardizing my variables using zval so that they have a mean of 0 and a variance of 1. For SEM, I just read, that observed variables are assumed not to have means of 0. Does this mean I should leave my variables in their original form?

zval has left the standardized variable with means not equal but very close to zero. I am not sure if that makes any difference in this case.

Thank you in advance.


Stata 15(64 bit) on windows 10. n=500 from survey data.

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

SEM works fine with non-standardized variables. Generally, folks on this list prefer to work with unstandardized variables since they are easier to interpret.

[Julie Janssens]

@Victoria: did you find an answer to your question yet? I have the same problem... Not sure whether I have to standardize variables or not. I didn't even know you had to standardize before doing factor analysis. Is this always required? And how do you use binary variables in factor analysis, I thought you at least needed ordinal variables? Thank you for your help in advance! Julie

[Victoria Lorvig]

Phil Bromiley Thank you for your response. I will probably switch back to standardized for my final regressions.