| Message/Author |
|
|
|
|
Is there anyway to specify squared multiple correlations (instead of ones) as priors when conducting an EFA? If so, what is the code?
Thanks! |
|
|
| bmuthen posted on Tuesday, October 11, 2005 - 4:31 pm
|
|
|
| The use of priors for the diagonal of the correlation matrix for EFA is particular to older EFA estimators (such as principal factoring). Mplus does not use those techniques. Instead, the default estimator is unweighted least squares, and maximum-likelihood can also be requested. |
|
|
| Anonymous posted on Wednesday, October 12, 2005 - 12:58 pm
|
|
|
Hi,
When using WLSMV estimator for an EFA is there a way to aid in determining the number of factors by 1) either examining the probability value for the Chi-Square test (non-significant) or 2) conducting a Chi-Square difference test from 2 competing number of factors?
Thanks in advance! |
|
|
|
|
| You will obtain a couple of fit statistics with the WLSMV estimator that you can use to decide on the number of factors. You will also obtain eigenvalues and residual variances. |
|
|
|
|
Hello Drs. Muthen,
It is quite common that researchers often report standardized loadings when conducting CFAs. Could that information be used somehow as priors using Bayes? If so, how would the command read:
a1-b ~N(0, .###)
thank you. |
|
|
|
|
| Priors should be in the same scale as the scale of the raw parameter estimates. They should be determined from an independent source not the data being analyzed. |
|
|
| Back to top |
