missing values

[bkoksal]

Does garch handle missing values?

For return series, the values for the weekends are missing. Should I remove those lines before I process the data? Thanks.

[TomDoan]

Does garch handle missing values?

No.

For return series, the values for the weekends are missing. Should I remove those lines before I process the data? Thanks.

You probably want to do business day data rather than seven day a week data. But if you also need to skip non-trading holidays, you would need to squeeze out the missing values.

[bkoksal]

I want to estimate a multivariate garch model for NYSE and London Stock Exchange. Removing the weekends is easy. But non-trading holidays are different for both countries. How can I deal with this problem for multivariate garch? Should I just remove the days for "both" countries when return data are missing for "one" of the countries because of a non-trading holiday?[/quote]

[TomDoan]

I want to estimate a multivariate garch model for NYSE and London Stock Exchange. Removing the weekends is easy. But non-trading holidays are different for both countries. How can I deal with this problem for multivariate garch? Should I just remove the days for "both" countries when return data are missing for "one" of the countries because of a non-trading holiday?

I would think that simply removing the returns for both countries when either is missing is the worst of the three possible ways of handling this, since you would be leaving out possibly very relevant data. Either plugging in a zero return for the non-trading country (which seems to be the most common way of handling this that I've seen), or dropping the data point but computing returns for both series over the two days would be better.

[jack]

Dear Tom'

Suppose that I have following data with a missing value (d) and I want to compute returns for them. Is it correct to proceed this way?: squeezing out the "d" and computing return between "e" and"c"? (ln(e/c)*100).

a. 2
b. 4
c. 5
d. NA
e. 7
f. 8

[TomDoan]

As I said above, there are two ways to handle this and both get used. With a single series, it would typically be to drop D entirely.