Message/Author |
|
EmilyBo posted on Tuesday, February 21, 2017 - 3:26 pm
|
|
|
I would like to fix the scale of the results and in order to do that I would like to fix the latent class pattern probabilities. In other words, I used a much bigger data set to obtain the latent class probabilities (the last column shown below) FINAL CLASS COUNTS AND PROPORTIONS FOR THE LATENT CLASS PATTERNS BASED ON THE ESTIMATED MODEL Latent Class Pattern 1 1 2.53840 0.06044 1 2 0.00000 0.00000 1 3 12.41348 0.29556 2 1 4.59199 0.10933 2 2 0.00000 0.00000 2 3 22.45612 0.53467 I would like to fix those value in another round of latent class analysis. I would appreciate any insight on the issue. |
|
|
You fix them by fixing the latent class logits - see UG chapter 14 pages 500-503 for 3 representations of how the logits influence the joint probabilities. |
|
EmilyBo posted on Wednesday, February 22, 2017 - 2:43 pm
|
|
|
Okay.Thank you, Bengt. Would you please take a look at my code below? I ran into error messages and I am not sure what went wrong. Am I correct that I can fix the logit of latent class probabilities? The latent class probabilities are shown below. FINAL CLASS COUNTS AND PROPORTIONS FOR EACH LATENT CLASS VARIABLE BASED ON ESTIMATED POSTERIOR PROBABILITIES Latent Class Variable Class C1 1 81142.20312 0.27499 2 213931.79688 0.72501 C2 1 57923.91016 0.19630 2 118573.26562 0.40184 3 118576.82812 0.40185 In the code below I fix C1#1 to be -0.42103, which is logit(0.27499). Am I on the right track? Model: %overall% model c1: %C1#1% [C1#1@-0.42103]; [und@-0.714 met@-1.450]; und@0.817 met@0.252; %C1#2% [C1#2@0.421027]; [und@0.159 met@0.637]; und@0.817 met@0.252; ..... |
|
|
Note that several logits, not just one, play into the probability as shown on those pages in the UG that I mentioned. If this doesn't help, we need to see the full output - send to Support along with your license number. |
|
Back to top |