Number of decimals PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Trang Q. Nguyen posted on Tuesday, November 12, 2013 - 3:31 pm
Could you please tell me if there is a way to ask Mplus for parameter estimates with more than 3 decimal places? I need to use regression coefficients from a latent class regression to calculate individual level class membership probabilities and ratios of probabilities, while fixing some covariates and keeping some unchanged. Just to check if things run OK, I calculated predicted probabilities using actual covariates from data and averaged them to recover class prevalence. These are close to but do not match model estimated class prevalence (see below), so I realized this is perhaps due to rounding of parameter estimates.

class prevalence from Mplus output:

Latent Classes

1 108.02221 0.04393
2 166.61008 0.06776
3 465.14055 0.18916
4 813.52596 0.33084
5 905.70120 0.36832

class prevalence derived from regression coefficients:

> mean(p1)
[1] 0.04479309
> mean(p2)
[1] 0.07153347
> mean(p3)
[1] 0.1762071
> mean(p4)
[1] 0.3363368
> mean(p5)
[1] 0.3711296

Thanks much for your help!
 Trang Q. Nguyen posted on Tuesday, November 12, 2013 - 4:16 pm
I found a mix-up in the order of my covariates. After fixing that, derived class prevalence is much closer to Mplus output:

0.04395251
0.06780296
0.18910248
0.33077608
0.36836597

which works for my current analysis. For the future, though, I would appreciate knowing if there is a way to get estimates with more decimal places.

Thanks much!
 Linda K. Muthen posted on Tuesday, November 12, 2013 - 5:22 pm
Use the RESULTS option of the SAVEDATA command.
 Trang Q. Nguyen posted on Tuesday, November 12, 2013 - 6:12 pm
Thank you so much!
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: