| Bootstrap
Standard Errors - Methods |
| In order to
provide researchers more information about the estimates reported in our LIS
Key Figures Series, we have computed bootstrap standard errors
for several of our figures. For those of you who are interested in our
methods, we used Stata 7.0 and have provided the programs on our page.
In order to compute 95% confidence intervals based on normal
distributions, multiply bootstrap standard errors of the estimates by
1.96. |
| The Stata manuals
provide nice overviews of the bootstrap technique (see #8, pp. 79-88), as
do the other references cited below. While the procedures we use are now
well-established, there is still NO fixed rule concerning the number of replications one should use in
computing bootstrap standard errors. In fact, since we are only reporting
standard errors and recommending the calculation of confidence intervals
based upon the normal distribution, 50-200 replications should be
sufficient (see Ref. #8, p. 81 and Ref. #2). However, in theory, an
infinite number of replications is desirable and thus the decision is
often a practical one. Therefore, in this case, we used 500 replications
of the bootstrap. Please note that maximum number of replications we will allow in remote
access is 1000. Quite simply, more replications interfere with the job
submissions of other researchers using the LIS. |
|
Suggested Readings: |
-
Efron, Bradley and Robert J. Tibshirani. 1993.
An Introduction
to the Bootstrap: Monographs on Statistics and Applied Probability,
Vol. 57. NY: Chapman and Hall.
-
Gould, William and Jeff Pitblado. 2001. “Guidelines for
bootstrap samples.” Stata
FAQ Statistics.
-
Heinrich,
Georges. 1998. “Changing Times, Testing Times: A Bootstrap Analysis of
Poverty and Inequality Using the PACO Data Base.” PACO Research Paper n°24.
-
Jäntii,
Markus and Sheldon Danziger. 2000. “Income Poverty in Advanced
Countries.” In Handbook of Income Distribution, Vol. 1. Atkinson,
A.B. and F. Bourguignon (eds.) Oxford: Elsevier.
-
Mills, Jeffrey A. and Sourushe Zandvakili. 1997. “Statistical
Inference via Bootstrapping for measures of Inequality.” Journal of
Applied Econometrics 12: 133-150.
-
Osberg,
Lars and Kuan Xu. 2000. “International comparison of poverty intensity:
Index decomposition and bootstrap inference.” Journal of Human
Resources 35(1): 51-81. [Also available as LIS Working Paper
#165].
-
___________.
1999. “Poverty Intensity- How do Canadian Provinces
Compare?”.
Luxembourg Income Study Working
Paper Number 203. Center for Policy Research, Syracuse University.
-
StataCorp. 1997. Stata Statistical Software: Release 5.0.
College Station, TX: Stata Corporation.
|
