Announcement

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

  • Diebold-Mariano test - Sensible?

    Hello guys,

    I am currently doing some research on forecast/prediction accuracy. My research design is straight-forward: I use two different models to predict a liquidity measure (out-of-sample). I use the previous 100 days to come up with a prediction for both models.

    However, I want to predict for only one day per firm in my sample. Thus, I have 100 rows, with one observation for each firm. I have three variables: actual_liquidity, liquidity_pred1, liquidity_pred2. I want to use the Diebold Mariano test.

    First, I wanted to use the following code:


    dmariano liquidity liquidity_pred1 liquidity_pred2, crit(MSE)

    But it resulted in the following error: "sample may not include multiple panels".


    I used a work-around, but I am totally unsure whether this is a sensible approach, since it doesn't seem clean.

    xtset, clear
    gen ts_var = _n
    tsset ts_var
    dmariano liquidity liquidity_pred1 liquidity_pred2, crit(MSE)

    As a sanity check, I variated the order of my observations before generating the ts_var, and it didn't change anything about the results.


    I would be very grateful if somebody could help me with my issue.



    Thank you in advance and best regards,
    Jonathan


  • #2
    Hey,

    does anybody have an idea? I am really unsure and grateful for every reply. If information is missing to understand my situation, please don't hesitate to let me know.


    Best regards,
    Jonathan

    Comment

    Working...
    X