Page 1 of 1

2-pass regression with time series data

Posted: Thu Oct 02, 2014 4:17 am
by fan
Hi Tom,

I am trying to apply 2-pass regression with time series data I have. I would like to how to use the coefficient estimates from my first equation as my regressors in my second equation. Many Thanks

Re: 2-pass regression with time series data

Posted: Thu Oct 02, 2014 11:24 am
by TomDoan
Isn't that the same basic idea as Fama and Macbeth? Just save the coefficients into series and use them as data in the second regression.

Re: 2-pass regression with time series data

Posted: Thu Oct 02, 2014 4:52 pm
by fan
TomDoan wrote:Isn't that the same basic idea as Fama and Macbeth? Just save the coefficients into series and use them as data in the second regression.
Thank you, Tom. I tried as you suggested here. but my code did not work. Could you please kindly take a look my code?

Code: Select all

linreg y
# constant x
compute beta1=%beta(1), beta2=%beta(2)

linreg y
# beta1 beta2

Re: 2-pass regression with time series data

Posted: Thu Oct 02, 2014 5:42 pm
by TomDoan
Don't you need to run separate regressions over t for each SB (whatever the latter index set is) and collect the coefficients from those?

Re: 2-pass regression with time series data

Posted: Thu Oct 02, 2014 7:03 pm
by fan
TomDoan wrote:Don't you need to run separate regressions over t for each SB (whatever the latter index set is) and collect the coefficients from those?
Thank you for your quick reply. I thought I could have the coefficients for each SB from my first regression.Is it not right?

Code: Select all

linreg sb1
    # constant x
    compute beta1=%beta(1), beta2=%beta(2)

    linreg sb1
    # beta1 beta2

Re: 2-pass regression with time series data

Posted: Thu Oct 02, 2014 8:02 pm
by TomDoan
No. One LINREG does one regression. You need a whole collection of regressions at the first stage.