[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
R: st: Comparing non-parametric bootstrap vs. Monte Carlo
Dear Professor Yaffee,
thanks a lot again.
[mailto:firstname.lastname@example.org] Per conto di Robert A Yaffee
Inviato: luned́ 14 maggio 2007 14.31
Oggetto: Re: st: Comparing non-parametric bootstrap vs. Monte Carlo
A reading of Kish, Lohr, Deming and others on sampling will show
differences due to finite population
effect will diminish as the sample size increases. At the size you propose
to investigate, there
should be little difference between the methods in which you are interested.
You will find more
on this in Efron's works on the correct number of bootstrap trials to use.
Robert A. Yaffee, Ph.D.
Shirley M. Ehrenkranz
School of Social Work
New York University
2100 Linwood Ave.
Fort Lee, NJ
----- Original Message -----
From: Maarten buis <email@example.com>
Date: Sunday, May 13, 2007 6:02 am
Subject: Re: st: Comparing non-parametric bootstrap vs. Monte Carlo
> --- Carlo Lazzaro <firstname.lastname@example.org> wrote:
> > I would like to compare the results of a 10,000-size vector
> > non-parametric bootstrap simuation performed on the difference of
> > two samples of healthcare costs drawn at patient-level with the
> > results of a Monte Carlo simulation on the same dataset.
> > My main aim would be to detect, if any, the difference between
> > sampling with and without reintroduction.
> > Is there any way with Stata 9 to perform a non-parametric Monte Carlo
> > simulation or should I impose a given distribution (uniform? log
> > normal?)on the original data set before starting the simulation?
> I understand from your question that you want to compare the bootstrap,
> which draws random samples from your data with replacement, with
> another Monte Carlo simulation which draws from a non-parametric
> representation of your data without replacement.
> What makes the bootstrap non-parametric is that it immediately draws
> from the observed data, so it won't need a model (with parameters) to
> draw its replications. The data itself is a non-parametric
> representation of itself (and hopefully the population). In order to
> get estimates for standard errors you need to make sure that the
> bootstrap replications have the same number of observations as the
> original data. In your Monte Carlo implementation you would draw
> samples from your data without replacement. In that case you would
> reproduce the original dataset every time you draw the sample. So that
> won't work. In other words, if you want to sample without replacement
> you will have to impose a model.
> Hope this helps,
> Maarten L. Buis
> Department of Social Research Methodology
> Vrije Universiteit Amsterdam
> Boelelaan 1081
> 1081 HV Amsterdam
> The Netherlands
> visiting address:
> Buitenveldertselaan 3 (Metropolitan), room Z434
> +31 20 5986715
> New Yahoo! Mail is the ultimate force in competitive emailing. Find
> out more at the Yahoo! Mail Championships. Plus: play games and win
> * 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/
* For searches and help try:
* For searches and help try: