Firstly I am a newbie so I will apologise in advance if this is in the wrong forum. My question concerns testing for an equity trading strategy and I would appreciate any comments as to whether my approach is sensible or an alternative methodology would be more appropriate.

(a) Since I have a limited amount of historical data, circa 10 years, for test purposes and to reduce data snooping I have selected to use a stationary bootstrap technique to resample the returns data.

(b) I then evaluate the various strategies based on the resampled data.

(c) Based on repeating steps (a) and (b) for say 100 iterations I then determine the modal value of the test metric.

(d) Thereafter I carry out a paired 'T' test on the modal value to identify whether any variation is significant.

I would very much appreciate guidance as to whether the above approach is sensible.



| improve this question | | | | |
  • $\begingroup$ Something puzzles me. The idea of using bootstrapping has merit, it is better than what most people would do, which is the T test. But then, suddenly in Step (d) you revert to a T test, I am not sure why. Usually in bootstrapping I find out how the results I obtained rank among the bootstrapped results. If they rank in the top 5% of bootstrapped results, they are significant. No T test needed. $\endgroup$ – Alex C Jan 17 '16 at 22:01
  • $\begingroup$ Sorry Alex but the methodology I described was not entirely accurate and you should ignore the comment about modal value. I very much like the simplicity of your approach of just selecting the top X% values from the bootstrap results. Also thank you for confirming the suitability of the bootstrap. - Chris $\endgroup$ – Chris Wilson Jan 24 '16 at 19:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.