|
|
Discrepancy in number of observation |
|
Message/Author |
|
|
Hello, I am working on a latent class growth analysis with complex survey weights. There is a discrepancy in the final class counts for the latent classes based on their most likely latent class membership between the numbers presented in the report and when I output the data into a text file and then count the data using STATA. However, there is no difference in the proportion. Please see below for the relevant output in STATA and MPLUS. MPLUS Output: Class Counts and Proportions Latent Classes 1 2204 0.17842 2 4162 0.33695 3 1534 0.12423 4 4451 0.36040 STATA Output: ---------------------------------------------------------- Class | percentage lb ub obs ----------+----------------------------------------------- Presiste | 17.84 16.24 19.56 2557 Low-Unhe | 33.7 32.6 34.81 4421 Healthy | 12.42 11.58 13.32 1308 High-Unh | 36.04 34.73 37.37 4065 | Total | 100 1.2e+04 ---------------------------------------------------------- |
|
|
The Mplus output is weighted by the survey weights. If a person is classified to be in class 1, we count that person not as 1 but as the standardized sampling weight. The final number is rounded to an integer. If that doesn't resolve the issue, send your example to support@statmodel.com |
|
|
Thank you for that clarification. |
|
Back to top |
|
|