Message/Author 


I have the following code for a dataset I am running. VARIABLE: NAMES ARE ID GP Y1Y18 U1U3 MFS OVS AMS WVS LCS OV RLN SM WV LC RDN; USEVARIABLES ARE Y1Y18 U1U3 OV SM RDN; CLASSES = cg (4) c(2); KNOWNCLASS = cg (GP = 1 GP = 2 GP = 4 GP = 5); CATEGORICAL = U1U3; CENTERING = GRANDMEAN (OV SM RDN); MISSING IS BLANK; AUXILIARY = ID; ANALYSIS: TYPE=MIXTURE MISSING; ESTIMATOR=MLR; STARTS = 20 2; MODEL: %OVERALL% i s q Y1@0 Y2@1 Y3@2 Y4@3 Y5@4 Y6@5 Y7@6 Y8@7 Y9@8 Y10@9 Y11@10 Y12@11 Y13@12 Y14@13 Y15@14 Y16@15 Y17@16 Y18@17; i s q ON OV SM RDN; f BY U1U3@1; c#1 ON cg#1 OV SM RDN; c#1 ON cg#2 OV SM RDN; c#1 ON cg#3 OV SM RDN; My question is whether there is a way to specify just 1 latent class in cg#3? Thanks, Don 


cg#3 refers to class 3 of the categorical latent variable cg. If you want the categorical latent variable cg to have one class, you would specify CLASSES = cg (1); 


Yes I understand that, however wouldn't that set the number of classes to 1 in all known groups? What I was hoping to achieve was cg#1(2), cg#2(2), cg#3(1), and cg#4(2). Can this be done within the classes command? Thanks, Don 


I think what you are asking is whether a known class can have a different number of classes on the other categorical latent variable. I think you would specify this as follows: MODEL: %OVERALL% c#2 ON cg#3@15; 


Apologies for the simplicity of this question, but I just need to clarify.... I am running a growth mixture model with known classes (multiple group analysis), using a binary dependent variable U. In the output there is a threshold value (identical across groups), as well as an intercept and slope that are unique to each group. The probability of U = 1/(1+e(threshold + intercept + slope*time)) Logit = threshold + slope*time Correct? 


The logit is your argument: threshold + intercept + slope*time Note that this gives the probability conditional on these intercept and slope values. 

C. Lechner posted on Wednesday, October 10, 2018  4:22 am



Dear Linda and Bengt, I am trying to estimate a multiplegroup LCA model in which the number of classes differs across groups (i.e., known classes, cg). I have twelve groups and up to five latent classes: classes = c(5) cg(12). The model is supposed to be a fiveclass model for one group, a fourclass model for most other groups, and a threeclass model for two groups. I define the group with the highest number of latent classes (5) as the reference group by listing it last in the KNOWNCLASS option. When I constrain only one class proportion, this works fine: %overall% c#1 ON cg#1cg#11@15; The latent class probabilities of c#1 are fixed to zero in all groups save the reference group. However, when I add the second constraint, the latent class proportions come out unconstrained. In fact, only one class in the reference group is constrained to a zero proportion. %overall% c#1 ON cg#1cg#11@15; c#2 ON cg#1cg#2@15; I have tried various variants of this code, including separate statements for each group in cg, yet none has succeeded. Can you please point out the correct way of implementing the constraint? Thank you very much in advance. 


Please send your output and data if possible to Support along with your license number. 

Back to top 