Sanjoy posted on Wednesday, May 25, 2005 - 7:22 pm
Dear Professor ... some very basic queries needed to be cleared, I'm going through your student Yu's dissertation .. the problem arises because of the differences in nomenclature that's being used in econometrics vs. psychometrics
from Page 11 ...
Q1. "our hypothesized" model ...Does it mean the model we estimate at very FIRST/begining of our analysis ?
Q2. I suppose We also refer "our hypothesized" model as "UNrestricted" model ... right!
Q3. we refer a model as " more restricted" model when we put some restriction/s ON the model we FIRST estimate ...right! ...why do we use the word "more", does it have some special significance ... (in econometrics we just say either restricted or unrestricted)
Q4. What is the "Saturated Model" then ???
from Page 13(2.1.2) ....
Q5. again "more restricted baseline model" ... does it mean a model where we assume all the coefficients are zero !!! ...(usually in SAS, where it's feasible, they do this and give us an overall "F-fit" statistics and NOT chi)
Now from the Mplus output (I'm using WLSMV)...at the begining we get this " TESTS OF MODEL FIT Chi-Square Test of Model Fit Value 95.074* Degrees of Freedom 68** P-Value 0.0168 " Q6. What does it mean ... I mean, in comparison to what it's fitting something, or does it say that sigma(parameter hat) is pretty close to sample covariance
now just below ... we get another output
" Chi-Square Test of Model Fit for the Baseline Model Value 405.115 Degrees of Freedom 95 P-Value 0.0000 "
Q7. iS it the value when we assume all coefficients are zero
Q8. this difference is extremely statistically significant ....so we can reject that "all coefficients are zero" ..is it!
Given that you ask so many basic questions that can be answered by the standard literature, I will have to refer you to SEM texts such as Bollen's book.
Sanjoy posted on Thursday, May 26, 2005 - 12:02 pm
Thank you madam ... I'm going to get it and read.
Joyce T. posted on Thursday, July 14, 2005 - 7:46 am
Dear Dr. Muthen,
1- I would like to know if we can get and how, when using the MLMV command, the tests of model fit (which include Chi-Square Test of Model Fit, Chi-Square Test of Model Fit for the Baseline Model, CFI/TLI, Loglikelihood, Information Criteria, RMSEA, SRMR) that appears only in the output when the ML command is used.
2- I would like to know too, how in a structural equation model, can I declare that a variable is observed and endogenous.
You should get these fit statistics. If not, you need to send your output, data, and license number to firstname.lastname@example.org.
It is not necessary to declare that a variable in observed and endogenous. Mplus knows it is observed if it is part of the data and not defined by BY. It knows it is endogenous by the role it plays in the MODEL command.