Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Is it appropriate to use multilevel regression models for forecasting purposes?

    I have time series data about 10 individuals. For each of them I have 120 repeated measures over time (a monthly measure for 10 years).
    I have a dependent variable (Y) and several predictors (X1, X2, ...).

    So far I used multilevel models -mixed- to see which predictors were significant in explaining Y.
    Now I would like to use the model for forecasting purposes.

    Is it appropriate to use -mixed- for forecasting purposes?
    Can you point me to a tutorial in stata to learn how to do that?
    Can you also point me to some literature that supports the use of multilevel regression models for forecasting purposes?

    Alternatively, which other approach would you reccomend? ARIMAX? or others?

    Thanks a lot!




  • #2
    You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

    In the manuals, after every Stata routine is a postestimation documentation section. It will tell you how to do predictions after mixed. One tricky issue is how you handle the levels or fixed effects - it is hard to predict outside the sample when the estimates depend on fixed effects based on individuals in the sample.

    Comment


    • #3
      The individuals for which I need to make the prediction are the same of the training sample. They do not change. For this reason I was wondering if mixed is appropriate for forecasting.
      The idea would be to train the model on 80% of my data and then use the other 20% for out of sample predictions.
      Can you give me an example of how to do that?

      Comment

      Working...
      X