Anonymous posted on Sunday, November 27, 2005 - 8:38 pm
I'm writing a paper that deals with LCA's with a CFA. I'm looking for paper(s) that have used this approach. I'm finding it rather simple to explain the LCA process, but not as easy to explain the process with a latent variable developed in a CFA process. In other words, I performed example 7.17 with my data and I want to ground this in some type of literature. Thanks in advance
ex 7.17 is a CFA mixture model. Bengt is currently writing several papers on this topic and you may want to send him an email about this to request a relevant paper - if this is understanding your question correctly.
Anonymous posted on Monday, November 28, 2005 - 6:52 pm
Thank you. I have just a few more questions. I just finished reading Lubke and Muthen (2005). So, now I'm interested in developing my step-by-step plan. In my case, I have one latent variable that is indicated by 7 observed variables that I want to examine for heterogenity. If I'm understanding the step-by-step plan correctly, we are to develop 3 models. Model 1 no restrictions on the model at all across the classes. Model 2 restrictions on the intercepts only across the classes. Model 3 restrictions on the factor loadings, intercepts, and residual variances across the classes. This will allow us to better understand the heterogenity of the latent variable across classes. Is this correct? Or am I missing something?
My concern is writing this for a nontechnical audience. That is, at the end of the day what do we have? one, two, or three classes?
These steps would apply to continuous factor indicators. I would do all parameters free, factor loadings equal, factor loadings and intercepts equal. Not all disciplines would require residual variances to be held equal for measurement invarinace. To test heterogeneity, you can test the equality of the factor mean and variance.
Jamie Vaske posted on Wednesday, September 22, 2010 - 9:13 am
Hello, I am estimating a second order latent class growth analysis, where the first order latent model is a CFA of three factors ("measured by" three indicators) at three time points, and then the second order latent model is an LCGA of those factors. I have successfully established measurement invariance and estimated the model, but I am having trouble graphing the results.
If I use the TYPE = PLOT3; SERIES RISK94 RISK96 RISK98(*) or SERIES RISK94(0) RISK96(1) RISK98(2), I receive the error: "ERROR in PLOT command Time points for process 1 are not all continuous, all categorical, or all latent as they should be." (RISK94-98 are the names of the latent factors);
Is there something special I have to do in the SERIES command to get the observed and estimated means for the factors? I've also tried leaving off the (*) and (0, 1, 2) and I will not get the option to look at means.
Hello, I am conducting a Latent Class Analysis with random intercepts (which allow the mean on all my items to be different across individuals and in a way accounts for acquiescence). I am not sure if it works exactly like method factor in CFA. So I made two different models:
(Model 1) USEVARIABLES = v1-v10; CATEGORICAL ARE v1-v10; CLASSES = c(3); ANALYSIS: TYPE = MIXTURE; MODEL: %overall% F1 BY v1-v10@1;
(Model 2) the same, just added a new line in MODEL: F1 with c@0;
Why the second model asks for numerical integration? Which model should be used to account for random intercept? If the model 1 is used does it really mean that F1 is allowed to correlate with c?