Message/Author |
|
|
I am performing exploratory LCA and want to assign class memberships to each individual in my dataset to use a regressors in further analysis. I've heard of a method somewhat like multiple imputation for assigning membership, which may reduce the uncertainty involved in assigning membership. Is MPLUS capable of assigning class membership using an imputation idea? What are the pros/cons of using this method? Do you have any references in this topic? |
|
bmuthen posted on Wednesday, May 11, 2005 - 8:17 am
|
|
|
I would suggest instead doing this in a single-step analysis where you include the variables of the "further analysis" in the model. You find applications of this approach in several of my papers in the context of predicting "distal outcomes" such as high school dropout, school removal, and alcohol dependence. Mplus currently does not do multiple imputations of the class memberships, but that probably gives results similar to what I propose. |
|
canpei posted on Monday, October 27, 2008 - 8:40 am
|
|
|
Hi, Could you please let me know the title of your papers? Besides, I want to know whether the new version Mplus can tell me who is in each latent class, i.e. the membership of each student. Thanks, canpei |
|
|
For references, see this web site under Papers, such as in the category Growth Mixture Modeling - e.g. my 2004 chapter in the Kaplan handbook. For "pseudo-class draws" similar to multiple imputation, see the tech appendix section and the document "Equality tests of means..." Membership in classes is obtained when requesting cprobabilities (see UG). |
|
|
Hi, in Latent Class Growth Analysis how do I treat cases that haven't been assigned to any class? Thanks, Giovanni |
|
|
All cases have posterior probabilities. What do you mean? |
|
|
Hi, thanks for your response. My problem is that when I save the file with the probabilities of association to each class, about 10% of my sample is missing while it was included from the beginning. My sample consists of more the 10,000 subjects. Giovanni |
|
|
my syntax: This is my syntax: Data: File is 2010_2013_no_missing.dat ; Variable: names = ID PB030 PH010 B_PH010 C_PH010 D_PH010 PB140 PB150 PB190 PE040 AGE AGE_CAT2 RB040 HS040 HS050 HS060 HS120 B_HS040 B_HS050 B_HS060 B_HS120 C_HS040 C_HS050 C_HS060 C_HS120 D_HS040 D_HS050 D_HS060 D_HS120; usevariables = PH010 B_PH010 C_PH010 D_PH010; CLASSES = c(3); IDVARIABLE = ID ; ANALYSIS: type = mixture ; ESTIMATOR IS MLR; STARTS IS 100 20 ; ITERATIONS = 1000; CONVERGENCE = 0.00005; Model: %OVERALL% i s | PH010@0 B_PH010@1 C_PH010@2 D_PH010@3; i-s@0; output: sampstat STANDARDIZED (STDYX) TECH1 TECH11 ; plot: TYPE IS PLOT2; SERIES IS PH010-D_PH010 (S); type=plot3; series=PH010(0) B_PH010(1) C_PH010(2) D_PH010(3); SAVEDATA: SAVE = CPROB; FILE = LCGA3classes.dat; |
|
|
Send the relevant files and your license number to support@statmodel.com. |
|
Back to top |