Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: turning low frequency data into high frequency


From   Daniel Egan <dp.egan@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: turning low frequency data into high frequency
Date   Mon, 14 Mar 2005 13:51:37 -0500

I do have to wonder though exactly what the implication of this would
be, technically speaking (and as an economist):

Since the monthly data will not vary within a month, any analysis that
looks at correllations using daily data will be in effect (a better
statistician can correct me on this) comparing the monthly means of
the daily data with the means of the monthly data, but with more
observations.

Also, if you interpret the expanded monthly data as a monthly average
for example, there are problems explaining what you mean:
    
    a) if the monthly data is the predictor*: how can the monthly
average "affect" the resulting value on the first day of the month,
when the average hasn't existed (in its data form) yet? The converse
is true as well.

    b) Months with more days will be over-represented. As such, if the
monthly figure is (for example) total sales, months in which you would
expect higher sales (because there are more days to sell things) will
be over-represented, inflating estimates. The difference is one of 3
days (28 days versus 31 days) at maximum, which represents a 10%
increase (roughly) of selling time.

If the monthly data is a one-time sample, or represents an
end-of-month cumulative figure (such as total sales), similar problems
arise.


I would say that a better method is to transform the daily data into
monthly, in whatever manner Fabia finds appropriate. Unless she is
aiming at a technique such as Nick Cox hints at, in which case carry
on...

Just my thoughts.

Cheers, 

Dan

*Variable notations do not imply causation, just which side of the
equation they appear on.



Nick Cox escrito:
 
> Perhaps I should have underlined that
> economists do this ... with the aid
> of a so-called indicator series
> at the higher frequency. Thus any
> inferences, or implications, that this is "making it up"
> are going too far. Like many statistical
> procedures, it is a matter of doing the
> best you can with the data you have.
> 
> Nick
> n.j.cox@durham.ac.uk
>> 
> > I think Fabia really did want what she
> > asked for. Economists do this!
> 
> Fabia Carvalho
> 
> > > does anyone know if Stata can generate a
> > > > series of high frequency data (eg. monthly) from an
> > > > original low frequency one (eg. quarterly)?
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index