"LR test of model vs. saturated" in SEM

[Vincent Li]

Hey guys,

I use STATA to do the pathway analysis via SEM.
Below the result table, there is a note reporting "LR test of model vs. saturated: chi2(3) =".
What does it mean? Does the higher value of chi2 means fitness of the model or the lower value?

Thank you very much for advice or reference recommendation.

[Phil Bromiley]

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

The glossary for the SEM documentation says: saturated model. A saturated model is a full covariance model—a model of fitted means and
covariances of observed variables without any restrictions on the values. Also see baseline model.
Saturated models apply only to standard linear SEMs. See also SEM documentation example 1 and 4. A significant chi2 means we can reject the constraints that differentiate the estimated model from the saturated model - an indication of poor fit. Or as example 4 says with a statistically significant chi2: "The saturated model is the model that fits the covariances perfectly. We can reject at the 5% level (or any other level) that the model fits as well as the saturated model."

However, you can have a problem where a large sample makes it possible to reject many quite reasonable models.
This kind of issue is discussed in most standard introductions to practical SEM estimation.

[Richard Williams]

I recommend Acock's book for learning about sem basics and how to estimate models in Stata:

https://www.stata.com/bookstore/disc...g-using-stata/

[Vincent Li]

Thanks Phil.
I'll read more about SEM and learn about the examples to know more about how to explain the results and how to adjust the model.
If there are still questions, I'll post the code and results after adjusting the model. Thanks for your advice.

[Vincent Li]

Thank you Richard.
I have borrowed it from the library this morning and started to learn.
Thank you very much.