Hot answers tagged

6

I agree that this is confusing. Manifest variables and "observed" variables are both observed, in the sense that they are part of the input dataset, and not latent variables. The distinction made in the semopy package is that manifest variables are part of the measurement model (ie the definition of the latent variables), whereas "observed&...


5

Weighted Least Squares (WLS) is a general class of estimators where squared differences between observed and estimated values are weighted by some critereon. When the weights are all 1, this is Ordinary Least Squares. When the weights are based on unequal variances, this is Generalised Least Squares (GLS). For a specific choice of weights, based on the ...


3

You would normally choose GLS over ML for computational efficiency, or, with a very small sample, the biased-ness of ML. For severe departures from normality, you would not usually choose ML or GLS. The following table is taken from the seminal book by one of the godfathers of SEM, Kenneth Bollen, "Structural Equations with Latent Variables" , ...


3

Your model is empirically unidentified. This means that there is insufficient information in the data to estimate one or more of the latent variables. This could be simply due to a small sample size or that those latent variables are simply not measured well by the observed variables. In the model building process it is often a good idea to fit the ...


1

In short: Yes. Your models are not nested, so you should not be doing a chi-square test (using the anova() function) to compare them. (You can tell, partly because you have zero df for the difference). The AIC/BIC can be used to assess the difference in fit between the models, but because the df are the same, the difference in the AICs is the same as the ...


1

As it turned out, I misunderstood the concept of factor loadings. I tought that it's the correlation coefficient between latent and observed variable, but it's standarized regression coefficient instead and there is nothing wrong in value -1.17.


Only top voted, non community-wiki answers of a minimum length are eligible