|
|
Class prob. 0.000 and 1.000 for binar... |
|
Message/Author |
|
|
Hello, I'm running a 4-class model using 11 binary indicators. My sample is small, with only 240 cases with missing data. Everything looks fine, except that the class probabilities for some of the indicators were estimated as 1 and 0 (I have included below a portion of the output). Does this mean Mplus did not estimate these parameters and set them as fixed? if so, is this because the sample is too small? or could this be due to a different reason? I would appreciate any advice. Thanks! RESULTS IN PROBABILITY SCALE Latent Class 1 V42BR Category 1 0.305 0.166 1.835 0.066 Category 2 0.695 0.166 4.190 0.000 V42CR Category 1 0.302 0.185 1.635 0.102 Category 2 0.698 0.185 3.776 0.000 V42ER Category 1 0.587 0.140 4.202 0.000 Category 2 0.413 0.140 2.958 0.003 V30AR Category 1 1.000 0.000 0.000 1.000 Category 2 0.000 0.000 0.000 1.000 V30DR Category 1 1.000 0.000 0.000 1.000 Category 2 0.000 0.000 0.000 1.000 V30ER Category 1 0.758 0.144 5.268 0.000 Category 2 0.242 0.144 1.679 0.093 |
|
|
Having probabilities of 0 or 1 for some items is useful for the interpretation of the latent classes; these items are clear indicator of class membership. The 0/1 values are either from very large/small logits freely estimated or fixed because they were indeterminately large. This is not necessarily because the sample is small. |
|
Back to top |
|
|