For some reason this message does not appear on the digest.
Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html
Begin forwarded message:
From: Kit Baum <baum@bc.edu>
Date: May 15, 2006 12:09:59 PM EDT
To: statalist@hsphsun2.harvard.edu
Subject: re: high frequency data with panel structure
Sebastien said
Thanks for your reply. Actually the individuals are banks, hence to
incorporate
cash taker effects I wanted to run a panel regression to take in
this specific
effect. The data are interbank transactions - some are on a secured
basis
some are not. To look at the interest rate effect of different
collaterals
provided & maturity chosen I thought a panel data analysis to be
the best.
I already thought of changing the strucutre of the data into hours,
for instance.
Unfortunately I wasn't able to break it down to a lower level than
one day
in Stata.
This sort of data (e.g. trades, which may occur none, one or many
times per day in a given session) is not panel data. It is repeated
measures data. Think of an intensive care ward where some patients'
vital signs are monitored more frequently than others. We may have
50 BP measurements for patient A and 10 BP measurements for patient
B over the same calendar day. You do want to note that many of the
measurements come from the same unit (bank, patient etc.) but you
can do that with Stata's xt* commands using "iis". A command like
xtreg will allow you to specify i(banknr), and run a fixed-effect
model in which you demean each bank's transactions without the
necessity of evenly spaced time units from which the measurements
are generated. Of course, you could also add temporal variables to
the analysis--dummies for each trading day in your sample, or for
the day of week (Monday dummy, etc.)
Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html
*
* 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/