David Bard posted on Tuesday, December 12, 2017 - 9:55 am
Is it possible to get forecast estimates from an N=1 time series model? I tried to create the next observation in my time series and impute the missing value, but it appears that Mplus will not impute values at the beginning or end of a series. Is that correct? Is there another way to achieve this goal short of recreating the MCMC chains myself outside of Mplus? Here's the warning I get with the current approach:
*** WARNING Data set contains cases with missing on all variables. Cases that appeared before the first case with observed data and that appeared after the last case with observed data were not included in the analysis.
While not very elegant, either one of these will work:
1. instead of adding just one missing value at the end, add 20 more, and then copy over the last observation after the 20 missing observations. Because of the time distance of the 20 periods, the second copy of the last observation will have no influence (or very little influence on the model)
2. You can add a new dependent variable (for example the time variable) that is not missing in the beginning or the end of the series. To make sure that the new dependent variable doesn't affect your original model, specify the model for the new dependent variable to be independent of your original model.