13

I am learning about regression. I have done some cross sectional regressions which are fine. I recently did a simple time series regression. So I have a y & x vectors each containing 1000 observations.

I did a simple ols regression in excel which I thought was fine. However my online tutor e-mailed me after I sent the results asking what rolling window I was using? I don't seem to follow. I have e-mailed back but received an out of office reply and I'm keen to understand what he meant.

Is it simply rather than using all 1000 observations that I should use the first 50 observations (so 1:49) for a regression then do another regression using obs (2:50)?

My understanding was that as long as the data in my vectors is ordered correctly (in my case my date) then the regression should be done in the same way as a cross sectional regression - am I wrong?

mpiktas
  • 33,140
  • 5
  • 82
  • 138
mHelpMe
  • 627
  • 2
  • 10
  • 14

1 Answers1

20

Performing a rolling regression (a regression with a rolling time window) simply means, that you conduct regressions over and over again, with subsamples of your original full sample.

  • For example you could perform the regressions using windows with a size of 50 each, i.e. from 1:50, then from 51:100 etc.

  • Another approach would be to apply overlapping windows with a size of 50 each. So for example using 1:50, then 41:90 etc. (cutting off the last 10 elements in each succeeding subsample regression).

As a result you will receive a time series of your regression coefficients, which you can then analyze.

Which approach to follow depends a lot on your context and what your research is aiming for.

EDC
  • 378
  • 1
  • 3
  • 7
  • 1
    EDC provides a fine answer. I would usually align the window to a seasonal boundary, so each window would be equally affected by any seasonality. So, If I had weekly data, I might use 1-52, 14-65, 27-78, etc for overlapping windows. – zbicyclist Apr 11 '15 at 16:10