Begin forwarded message:
From: Kit Baum <[email protected]>
Date: May 11, 2004 4:52:10 PM EDT
To: "de la Garza, Adrian" <[email protected]>
Subject: Re: running mean and s.d.
Adrian,
the shortcut formula (which we used to use in the days of punch cards,
to avoid holding the entire dataset in 'core'!) is
s^2 = [ sum (x_i^2) - n (\bar{x})^2 ] / (n -1)
If you knew the mean, the \sigma^2 formula would have n \mu^2 in the
numerator and n in the denominator.
I believe that my shortcut s.d. formula works perfectly well for any
number of issues per issuer.
Please note that in the mail message the if _n == `i' clause appeared
on the next line. That is part of the bys borrname: replace sd
statement.
Thanks
Kit
On May 11, 2004, at 4:40 PM, de la Garza, Adrian wrote:
Chris,
Can you please explain your procedure? I see that you try to use the
standard deviation formula but I don't know how your forvalues loop
work, especially this part:
`i'*meanmat[`i']^2)/(`i'-1) if _n==`i'
what is that `i' multiplying meanmat[`i']^2?
Thank you.
Adrian
-----Original Message-----
From: Christopher F Baum [mailto:[email protected]]
Sent: Monday, May 10, 2004 9:05 PM
To: [email protected]
Subject: st: running mean and s.d.
Adrian asked how to calculate a running mean and s.d. within
bygroups: e.g.
borrower closedate maturity mean sdev cvar
18 Thermal Po~s 23-Dec-02 5 5.00 0.00 0.00
1st Silicon (~c 19-Jun-01 7 7.00 0.00 0.00
AB Capital Ma~d 25-Jul-97 2 2.00 0.00 0.00
AB Lietuvos D~a 25-Aug-00 3 3.00 0.00 0.00
ABC Roma Bank~d 13-Mar-97 1 1.00 0.00 0.00
ABN AMRO Bank~V 6-Apr-01 1 1.00 0.00 0.00
ABN AMRO Bank~V 9-May-01 1 1.00 0.00 0.00
ABN AMRO Bank~) 25-Oct-95 3 3.00 0.00 0.00
AC Internatio~d 30-May-96 6 6.00 0.00 0.00
AC Internatio~d 29-Nov-96 5 5.50 0.71 0.13
AC Internatio~d 16-Jul-99 6 5.67 0.58 0.10
ACINDAR Indus~A 16-Jul-93 2 2.00 0.00 0.00
ACINDAR Indus~A 12-Nov-93 5 3.50 2.12 0.61
ACINDAR Indus~A 14-Feb-97 7 4.67 2.52 0.54
ADI Cayman Corp 21-Dec-95 5 5.00 0.00 0.00
AECI Ltd 19-Jul-91 4 4.00 0.00
0.00
AIG Finance (~d 12-Apr-01 3 3.00 0.00 0.00
ANACAFE - Aso~e 31-Aug-93 5 5.00 0.00 0.00
AO IrkutskEne~o 20-Apr-98 3 3.00 0.00 0.00
AO Moscow Cit~k 19-Mar-98 3 3.00 0.00 0.00
AO Siberian O~) 15-Aug-97 3 3.00 0.00 0.00
APP Global Fi~ 17-Apr-97 5 5.00 0.00 0.00
APP Global Fi~d 4-Oct-96 5 5.00 0.00 0.00
APP Internati~V 18-Sep-95 5 5.00 0.00 0.00
APP Internati~V 18-Sep-95 10 7.50 3.54 0.47
APP Internati~V 28-Jun-96 3 6.00 3.61 0.60
APP Internati~V 12-Dec-96 3 5.25 3.30 0.63
bys borrname:g totmat=sum(mat)
bys borrname: g iss=_n
g meanmat = totmat/iss
g matsq = mat^2
bys borrname: g sqmat = sum(matsq)
g sd = 0
su iss,meanonly
local maxiss = r(max)
forv i=2/`maxiss' {
bys borrname: replace sd = (sqmat[`i'] -
`i'*meanmat[`i']^2)/(`i'-1) if _n==`i'
}
replace sd=sqrt(sd) if sd>0
g cv = sd/meanmat
l borrname issdate meanmat sd cv
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/
