1
$\begingroup$

I constructed the realized variance of bitcoin returns per day from 8-10-2015 to today. The realized variance is calculated by taking the cumulative squared intra-day returns. 5-minute high frequency data is used to estimate the realized variance. Daily realized variances are regressed on the realized variance one day prior and the average volatility over the past week and month.

However, there are days when there are many missing values. (this is if no action has been taken within 5 minutes). These are also the days when the number of trades is very low. Should I exclude all those days from my data set? So you should have 288 5-minute squared returns, some days have 200 or even lower. From which number should I exclude ?

$\endgroup$
1
  • $\begingroup$ What do you think about my answer? If it is helpful and clear, you may accept it by clicking on the tick mark to the left. Otherwise, you may ask for further clarification. This is how Cross Validated works. $\endgroup$ May 13 at 12:53

1 Answer 1

0
$\begingroup$

It might make sense to replace these missing values with zeros. When there are no trades, the price stays constant and so the variance is zero. For alternative solutions, you may also check how earlier studies on the same data have treated the problem.

In relation to your HAR model for daily realized variances, if you think days with few trades are exceptional, you may account for that by introducing dummy variables for such days or by other means.

$\endgroup$
2

Your Answer

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

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