+ Reply to Thread
Results 1 to 2 of 2

Thread: OLS with dummy and few observations (12 mean values from 2.5k observations)

  1. #1
    Points: 8, Level: 1
    Level completed: 15%, Points required for next Level: 42

    Posts
    1
    Thanks
    1
    Thanked 0 Times in 0 Posts

    OLS with dummy and few observations (12 mean values from 2.5k observations)




    Hi all

    Maybe you can help me with a problem I have in using OLS regression:

    I have a dataset of 12 markets, in total ~2.5k trades happened over all markets.

    I now calculated 6 different measures for the each markets performance (my 12 observations per measure i want to use as dependent variables)

    My treatment is indicated by a dummy, 6 markets were in one treatment, 6 in the other one.

    What I would like to do is, to use OLS to see, if there is a treatment effect, using the dummy variable, looking like this:

    Y = cons + b*dummy + e

    My question now is: Can i do this without overfitting? I dont know if I can use OLS because of my low number of observations, even though they stem from far more (and hopefully carry the statistical power / information of all my observations of transactions?)

    I would also like to add (2-3) control variables and am worried that I will use too many explanatory variables on too few observations.

    Thank you very much vor any help!

  2. #2
    Points: 3,457, Level: 36
    Level completed: 72%, Points required for next Level: 43

    Posts
    156
    Thanks
    42
    Thanked 43 Times in 40 Posts

    Re: OLS with dummy and few observations (12 mean values from 2.5k observations)


    Sounds like you have an N of 12. The number of trades doesn't seem relevant because you are saying things about the markets not thetrades but perhaps I misunderstood.

    The rule of thumb is that you want 10 cases for each independent or control variable so you could use just the dummy variable but not any explanatory variables. If you can't get more data, then one idea is matching the cases in one treatment to those in the other; there are various ways to do this, but, with low N, one good method is via propensity scores

  3. The Following User Says Thank You to PeterFlom For This Useful Post:

    callino (12-04-2014)

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats