|
Message/Author |
|
stats787 posted on Tuesday, June 14, 2011 - 8:23 am
|
|
|
Dear Dr. Muthen, I ran LCA for a set of 6 binary variables using 3-class model. I obtained the output, but I am not sure how do I group the variables based on the output. So based on the following output, can I assume latent class 1 contain V1, V2, V3, V5 and V6 since they have p-values < 0.05 in the last column. Thank you very much. P Latent Class 1 V1 Category 1 0.858 0.038 22.463 0.000 Category 2 0.142 0.038 3.710 0.000 V2 Category 1 0.935 0.027 34.526 0.000 Category 2 0.065 0.027 2.405 0.016 V3 Category 1 0.844 0.035 24.097 0.000 Category 2 0.156 0.035 4.441 0.000 V4 Category 1 0.855 0.074 11.601 0.000 Category 2 0.145 0.074 1.967 0.049 V5 Category 1 0.845 0.027 30.746 0.000 Category 2 0.155 0.027 5.650 0.000 V6 Category 1 0.944 0.025 38.329 0.000 Category 2 0.056 0.025 2.273 0.023 |
|
|
No, don't go by the p-values. The easiest way to understand which items measure which classes well is to plot the item profiles - that is, the item probabilities for each class. You can use the PLOT command of Mplus to do that. Also see the handout and video of Topic 5 on our web site regarding LCA. |
|
stats787 posted on Tuesday, June 14, 2011 - 6:19 pm
|
|
|
appreciated your advice. Thanks Dr. Muthen! |
|
Back to top |
|
|