1
$\begingroup$

I'm calculating the beta coefficients for some stocks using a single-index linear model with the OLS method. I'm computing the betas at different return intervals to assess the interval effect on the estimates. My question is, how do I handle outliers when the time horizon is 4 years? In some monthly returns (so around 50 observations), I observe significantly high kurtosis.

If you have an interesting paper about this topic, please tell me.

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ Chapter 16 of this book is entirely devoted to the topic. Apologies for the shameless plug. routledge.com/… $\endgroup$
    – BigBendRegion
    Commented Jan 13 at 12:01

1 Answer 1

Reset to default
1
$\begingroup$

Pinpointing your specific questions below:

My question is, how do I handle outliers when the time horizon is 4 years?

Generally speaking, outright removal of outliers is almost never a good thing, so using robust methods (including quantile regression, etc.) are generally preferred for overcoming any problematic observations. I'm sure you already know this, but because you are modeling time-based data, you want to make sure that is included in some way given this is an OLS regression, but that is more a secondary point.

If you have an interesting paper about this topic, please tell me.

The book linked here has an extensive section on this in Chapter 10, including detection and treatment.

Improve this answer
$\endgroup$
1
  • 1
    $\begingroup$ Thank you so much $\endgroup$
    – Mattia
    Commented Jan 12 at 7:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.