1
$\begingroup$

I am looking for a rigorous way to determine a suitable rolling window size for my stock data. Factors that will influence the window size are how fine my data is (minutely, daily, weekly etc.) and how quickly the underlying assets are moving (which can be analysed using the discrete Fourier transform). E.g. FX prices would fluctuate much faster than equity prices.

Can anyone point to papers, suggestions or strategies to deal with this problem?

$\endgroup$
  • 4
    $\begingroup$ What do you aim to achieve with the rolling window? What are you going to be putting the data into? $\endgroup$ – will May 30 '17 at 10:05
  • $\begingroup$ @will - I will be putting it into a transform such as discrete Fourier transform or PCA. $\endgroup$ – Chris B May 30 '17 at 11:20
  • $\begingroup$ What are the consequences of choosing "too long" or "too short" a window size? What are the tradeoffs involved? $\endgroup$ – noob2 May 30 '17 at 15:22
  • 1
    $\begingroup$ @noob2 - a "too long" window will not be sensitive to changes taking place at the moment. The consequences of a "too short" window is that it does not consider enough historic data and therefore is not particularly useful. $\endgroup$ – Chris B May 30 '17 at 17:31

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.