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

st: Bootstapping panel data

From   "M Parameswaran" <[email protected]>
To   [email protected]
Subject   st: Bootstapping panel data
Date   Tue, 28 Oct 2003 17:51:05 +0530 (IST)

Dear Frinds,

I am implementing a bootstrap estimation procedure using an unbalanced panel
data. The data set is very large and consists of 13016 observations on 1096
firms for 12 years.  In the bootstrap procedure I used cluster (block)
bootstrapping approach, i.e. treating observations on each firm as cluster and
sampling among clusters, available in STATA. The number of clusters in each
bootstrap sample is equal to the number of firms in the original data. However
of because of the unbalanced nature of the panel data the number of
observations is different in different bootstrap samples. (The difference is
not large. I found that the minimum number of observations in the bootstrap
samples is less than 150 of the original data and the maximum number of
observations in the bootstrap samples exceeds the number of observations in
original data by 152 in 300 replications.)

Is the any problem in bootstrapping in this way? Firms´┐Ż entry into and exit
from the data set is random.
If so, is there any STATA code available that can sample same number of firms
and same number of observations in a bootstrap procedure?

One pervious study, which used bootstrap estimation in unbalanced panel data
context, (Levinsohn and Pertin, 2003, Review of Economic Studies, pp. 317-341)
considers that a bootstrap sample is complete if it has same number of firms
and number of observations equal to or more than that of the original data.
Since, computationally it is quite difficult to fix the number of blocks and
the number of observations together in a bootstrap procedure, as an alternative
procedure, is there any problem in selecting only those bootstrap samples that
satisfy the criteria set up by the above mentioned study and estimates
parameters of interest using these samples. (In bootstrap samples around 50 per
cent of the samples satisfy this criteria)

Any comments on these issues would be greatly appreciated,
Thanking you in advance
M. Parameswaran

Parameswaran M
Centre for Development Studies,
Prasanthnagar Road, Ulloor,
Trivandrum 695 011,
Phone: 0091-0471-442481
E-mail: [email protected]
        [email protected]

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index