On Aug 12, 2004, at 2:33 AM, Jan wrote:
thanks you for your prompt responses. I understand your concern.
However, I am
applying -mvsumm- in a finance setting: I need to estimate the
volatility of
stock returns using the preceding 60 months of returns. Sometimes, I
have just
1 observation missing... and as long as this observation is part of
the
preceding 60 months, I get no estimate of the volatility for that
company's
stock (i.e. I lose 60 observations by insisting on a complete
history). I am
willing to make the tradeoff between gaining 60 observations at the
cost of a
slightly less reliable measure of volatility.
In your answer you suggested to clone and reprogram your original
-mvsumm-. I am
not a programming whiz - do you have any pointers?
This is a common procedure in finance - I think you may underestimate
the demand
for such a feature in -mvsumm- ;-)
Thanks a lot!! Jan
A reasonable approach would be to fill in the series with some form of
interpolation where you have missing observations in the middle of the
series. I don't quite understand how one can have one missing return
observation, since a return is a log price relative, and the prices
needed to estimate the missing observation will be used to compute the
prior and subsequent observations. But if for some reason you do not
have the original data available, then interpolation might be a
sensible option.
Kit
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
