+ Reply to Thread
Results 1 to 2 of 2

Thread: Difference in coefficients between regressions?

  1. #1
    Points: 2,848, Level: 32
    Level completed: 66%, Points required for next Level: 52

    Posts
    5
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Difference in coefficients between regressions?




    Hi - I'm looking for a way to test whether the beta coefficients are significantly different between two regression analyses. For example:
    Regression 1: Prejudice toward blacks is DV, ~30 sociological/social psych variables are the IVs
    Regression 2: Prejudice toward Asians is DV, ~30 sociological/social psych variables are the IVs

    Now, for example, I want to test whether the effect of an independent variable, say income, is SIGNIFICANTLY different between the first regression (blacks DV) and the second regression (Asians DV). How would that be done? The theory is interesting to me, but I especially need to know how to do this in Stata or SPSS (preferably Stata). I've read about the Chow test, but that seems useable only for regressions in which the DV is the same but in which you're testing the differences between subsets (e.g., the differences in the effect of the IVs on the DV between male vs. female respondents), not between regressions in which the DV is different.
    Thanks for any help.

  2. #2
    Points: 2,456, Level: 30
    Level completed: 4%, Points required for next Level: 144

    Posts
    5
    Thanks
    0
    Thanked 0 Times in 0 Posts

    this is a fun one


    i think it is kind of neat anyway, but then again, i am hanging out on a statistics board, so what do i know?

    If, and only if, your IV are the same, the only difference is your depvar response, then you do not need SPSS, just anything that can do a t-test.

    the coefficient is a correlation (partial) of X1 and Y conditional on all other X's. If all other X's are the same, then a t-test of differences will tell you whether Y1 (black) or Y2(asian) have a significant effect on this partial correlation Beta1.

    Any decent reg program can spit out the t-test means (the coefficients) and the standard errors that you need.

    Just get the t-test formula in front of you and fill it in with the coefficient data.
    Last edited by vinoverde; 03-25-2009 at 11:56 PM. Reason: too much info

+ Reply to Thread

           




Similar Threads

  1. Replies: 0
    Last Post: 11-29-2010, 10:30 AM
  2. Comparing coefficients for nested logistic regressions
    By dnh in forum Regression Analysis
    Replies: 9
    Last Post: 11-24-2010, 08:45 AM
  3. Test for significant difference in coefficients
    By mikethenerd2 in forum Psychology Statistics
    Replies: 2
    Last Post: 05-27-2010, 03:40 PM
  4. Replies: 2
    Last Post: 03-09-2008, 06:53 AM
  5. Replies: 4
    Last Post: 08-14-2007, 04:17 AM

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