Dear Dr.Austin Nichols and other users/experts,
I'm very sorry for not having informed my problem clearly. I shall do
it in this mail:
The data that I have is a survey containing monthly textile purchases
data from 1994-2003 for 20000 households. All households are not
repeating though, i.e., this is an unbalanced panel. In addition,
observations are literally absent for many months for many households
in each year. But the survey description mentions that the data was
filled only when there was some purchase, which was never a monthly
phenomenon in most cases, understandable for textiles in a developing
country like India, for which this data was collected. Hence, I feel
that it would not be a mistake considering the missing points to be
zero, as it directly follows from the way the survey is designed. But,
I was a little apprehensive about doing this, which is why I made the
contradictory statement.
I'm trying to estimate a demand system (Bias-corrected Dynamic
Linearised AIDS) using sureg in stata and hence I'm not using any
panel command other than informing stata that this is panel data,
using tsset.
Hope this info is sufficient. Please advise me on whether I should
fillin the data in this case, or can I use only the existing
observations?
From: austin nichols <[email protected]>
Subject: Re: Re: st: A query on handling monthly data
This is not a well-formed question, and contains two contradictory
clauses, viz. "it's implicit that the observations for which data is
not available should have been zero" and "it could as well have been
because of missing data!"
If the missing data are supposed to be zeros, fill in the zeros.
Otherwise leave them missing (note that -fillin- will add observations
with missing values). As for whether SUR will work (or is
econometrically legitimate) with or without imputing values, it is
impossible to know without knowing a lot about your research
project--and you have given no info at all on that.
In general, you don't need to eliminate the gaps in the data series
within each panel if you are going to pool your data and cluster on
panel ID (which offers the possibility of allowing for an arbitrary
within-panel serial correlation structure), but see the help file for
each of the relevant xt estimators (-help xt-). When you -tsset- the
data, Stata will tell you if you have gaps, and some estimators will
use panels with gaps in the series, and some won't.
As for the rest, you will have to come up with a series of more
specific questions, predicated on a careful but succinct description
of your data and research, if you want useful responses.
--
With Warm Regards,
Badri Narayanan G.,
Indira Gandhi Institute of Development Research,
Gen. Vaidya Marg, Goregaon (East),
Mumbai-400065
*
* 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/
