Message/Author |
|
|
The Tech 14 output from my latent class analysis says that I requested 250 bootstrap samples, and reports 250 successful bootstrap draws. However, the Tech 14 output contains a warning that says "The best loglikelihood value for the 7 class model for the generated data was not replicated in 11 of 17 bootstrap draws." I'm confused. Were 250 bootstrap draws actually completed or just 17? Why does the warning mention 17 draws when the output immediately above the warning says 250 bootstrap draws were successful? Thanks to anyone who can provide clarification! |
|
|
The successful bootstrap draws are 250. In 233 bootstrap draws the LRT is bigger than the observed LRT value (so the algorithm does not care if the likelihood was replicated, even if it was not replicated adding more random starts would make it only bigger which is not of interest because it is already bigger than the observed value and it would not affect the p-value computation in any way). Actually, the warning message is completely irrelevant in your case. Your p-value is already 0.93. Adding more random starts could potentially only increase this p-value but that has no impact on the inference because very clearly the 6-class solution could not be rejected in favor of a 7-class solution. |
|
|
Tihomir, Thank you for the explanation. Now I'm curious about something else. How did you know my p value was 0.93? I did not mention the p value in my post above, nor did I send Mplus my output, input, or data files. |
|
|
0.93=233/250 which is how the p-value is computed as we compare the bootstrap generated LRT distribution with the observed value. |
|
|
233 comes from 250-17 |
|
|
Thank you for your explanation again, Tihomir. Now I understand! |
|
Back to top |