Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: RE: Question on Stata Bootstrapping from PhD Student

From   Maarten Buis <>
Subject   st: RE: Question on Stata Bootstrapping from PhD Student
Date   Tue, 19 Feb 2013 10:39:37 +0100

--- Annas Abedin wrote to me privately:
> I am using Stata in my PhD research on consumer
> behavior and have seen a number of your postings
> on bootstrapping in Stata online.

This type of question should not be sent privately to a member of the
Statalist but to the Statalist instead. The reasons why that is very
important is discussed here:

> For a few regressions that form part of my research
> with a sample of ca. 300 consumers (1 categorical variable,
> 1 continuous from 1 to 5, 1 continuous from 1 to 7) the
> normality of errors and the homoscedasticity assumptions
> are violated, which is why I have to resorted to doing
> bootstrapped regressions as follows:
> ma789 g einkommen rrik if relwas++4, vce(bootstrap, reps(1000))

That does not look like valid Stata syntax. I suspect you wanted to type:

regress ma789 g einkommen rrik if relwas==4, vce(bootstrap, reps(1000))

> I wondered whether you could help me very briefly with respect to two things:
> 1) Would you agree that this kind of bootstrap is appropriate in my case?

There is no way for me to tell, that just depends on too many details.
If you are uncertain you will need to find an expert locally who is
willing to help you (possibly for a fee because you are asking a lot
of time from that expert), make an appointment with him or her, take
your data and .do-files with you, and spent time with that expert
going step by step through your research question, your research
design, your data preparation, and your final analysis. This is
something that cannot be done by email.

> 2) Could you refer me to a source that illustrates how I should report the
> results of the bootstrapping in my papers / in my thesis (or what output
> is critical to report)

I would report the results pretty much like the results of any other
estimation command. The only difference would be that in the titel to
the table (or graph) I would mention that the standard errors (or
p-value, or confidence intervals, or whatever you prefer) are based on
the bootstrap and mention the number of replications you used. When
you report confidence intervals I would also mention which type you

> and a source that explains what "type" of bootstrapping is performed by this function?

The helpfile and the manual entry for -bootstrap-, see -help
bootstrap- and -help bootstrap postestimation-.

-- Maarten

Maarten L. Buis
Reichpietschufer 50
10785 Berlin
*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index