We are testing a DSEM model which runs without the TINTERVAL option included, but crashes after all iterations are done when the TINTERVAL option is included. We tried it on multiple computers. There is an output file created, but it does not contain estimates. We have 228 participants who filled in questionnaires 5 times a day, with random time intervals, for 11 days. So max 55 measures over the measurement period. As time variable we use the minutes since the start of the measurement period (time_inc). We use a time interval of 2 hours (120). In the example below we used just 1000 iterations, but the same problem occurs when set to 50000 (which runs without TINTERVAL).
Are you familiar with this issue and do you know of a solution?
Model: %within% sSE | SE on SE&1; SE_REJ | SE on Reject&1; REJ_SE | Reject on SE&1; REJ | Reject on Reject&1;
SE_ACC| SE on Accept&1; ACC_SE | Accept on SE&1; ACC | Accept on Accept&1;
%between% SE Reject Accept SE_REJ REJ_SE ACC_SE SE_ACC on BSE; sSE with SE Reject Accept; SE_REJ with SE Reject Accept; REJ_SE with SE Reject Accept; REJ with SE Reject Accept; SE_ACC with SE Reject Accept; ACC_SE with SE Reject Accept; ACC with SE Reject Accept; sSE with SE_REJ REJ_SE REJ SE_ACC ACC_SE ACC; SE_REJ with REJ_SE REJ SE_ACC ACC_SE ACC; REJ_SE with REJ SE_ACC ACC_SE ACC; REJ with SE_ACC ACC_SE ACC; SE_ACC with ACC_SE ACC; ACC_SE with ACC; SE with Reject Accept; Reject with Accept; [BSE];
I am running a twolevel random DSEM model. I have daily measures of # of drinks per day for 30 days for 50 people. However, when I set my TINTERVAL for Drinks(1), my Mplus crashes. However, it works when I set it for Drinks (.9), but then I get 1650 observations with 150 missing instead of 1500 observations. Does this mean that I should not report the significant results that I get from this slightly different model? The .9 interval means that the number of observations per individual is 33 rather than 30.
As a follow up to the first post of the topic: We are running DSEM analyses with tinterval option. I used an increasing number of iterations to inspect the stability of the model. The model runs fine until 60900 iterations (PSR = 1.004), and gives output until then. However, after 60900 iterations I get the following error: THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY. THE ESTIMATED BETWEEN LEVEL POSTERIOR VARIANCE COVARIANCE MATRIX IS NOT POSITIVE DEFINITE AS IT SHOULD BE. THE PROBLEM OCCURRED IN CHAIN 1.
Do you have any idea how this could happen? Is this reason to be suspecious about the resuls with a lower number of iterations?
This can happen due to a variance going to zero - check the estimates in the output. If this is for a random effect, make it fixed instead. Also, look at the trace for the non-converging run to see which parameter has problems.