Hee-Jin Jun posted on Tuesday, October 23, 2007 - 11:05 am
I am working on trajectory analysis. Outcome variable is binary and we have 18 time point (age 0 to 17). Last three time points (age 15,16,17) we have some systematic missings (some kids haven't reached that age yet). First I did it with the fixed time points (0 to 17), but it doesn't seem to fit well. So I tried to add quadratic term. Unfortunately, it took quite a long time (few days) and then crashed (When I check the program, program was ended without output file.) Next, I freed the time points except the first two. One day later, I also had the same results (program ended with no output file).
I am wondeing if this is a known bug or new one for you and how can I get the results. Because there is no output file, I don't even get any error message.
Hee-Jin Jun posted on Tuesday, October 30, 2007 - 9:26 am
Running Mplus with a large datset with many time points has been challenging. While I am running a program with 12 time points (N=around 6300), I found that the running time has increased dramatically. E.g. one iteration took about 7500 seconds, which is about 2 hours, for a day since yesterday. I am wondering this is some kind of sign that the program has some difficulties, so that it's better to close and try it again. Thank you for your help.
e.g. 86 -0.60855213D+05 0.0010999 0.0000000 EM 7472.53 79127.1 87 -0.60855211D+05 0.0014535 0.0000000 EM 7457.41 184.5
Actually I may have got it to run but I have a follow-up question. My data is normally distributed which is why I originally went with ML but I got it to run using MLR and TYPE=RANDOM instead of ML. Is it ok to use MLR anyway in this context to get it to run?
I would want to understand why it did not run with ML. You may have different convergence criteria using RANDOM ML and MLR. You can check the outputs. It is never a good idea to leave an analysis without a clear understanding of the problem.
Yep that was my concern, I didn't know why it wouldn't run with ML. Ok, I will prepare the input and data to send to you and will try to write a bit about the steps I have/plan to take in case I have missed something important.
I have the same problem, I am trying to run a model with three continuous outcome variables using type = complex, but after about a minute Mplus stops running without giving any output. It does work when I lower the number of iterations, but then I get no convergence because the number of iterations exceeded. I am wondering why the model doesnt provide any output and whether there is something I can change to get Mplus providing output on my model?
I am trying to run a second order conditional latent growth curve using T-Scores. However, I've tried twice now and each time the program runs for a few hours before it seems to stop computing new iterations. It lists parameters and derivatives across the bottom of the DOS screen but DOS window doesn't close. It just seems to sit stagnant and I get no output.