| Message/Author |
|
|
|
|
what algorithm is used to estimate a standard mixture model w/ continuous indicators? it is EM, correct? i was just curious if it is the standard EM, because I noticed in the output it mentions "Optimization Specifications for the Quasi-Newton Algorithm" in addition to EM. Just looking for the relevent details to describe the method in a manuscript.
thanks |
|
|
|
|
| See the ALGORITHM option for a brief description of how this is done. |
|
|
|
|
| See the V5 UG, page 491 for a description (available on our web site). For mixtures, Mplus uses the "EMA" option, where "A" stands for acceleration. EMA combines EM steps with interspersed QN or FS steps when they are needed to speed things along. |
|
|
| John C posted on Thursday, April 20, 2017 - 7:06 pm
|
|
|
I'm conducting a latent class analysis and am having trouble with managing to replicate the best loglikelihood value.
I've tried several options including 100 sets of random values with 10 optimizations as in
ANALYSIS:
Type = Mixture;
Starts = 100 10;
The dataset is larger than any I've used in the past for LCA, with around 30,000 individuals. I'm also specifying around 8 different groups.
I'd appreciate any suggestions that might help with replicating the best log likelihood. |
|
|
|
|
| Send output to Support along with your license number. |
|
|
| Back to top |
