st: RE: Seeking info on CEM cutpoints algorithms - e.g., scott, ff, fd etc.

["Ben Hoen" <bhoen@lbl.gov>]

Thanks Matt.  I see the formulae for the various algorithms on page 24.

Also, I have quickly skimmed the document and found this discussion under
5.4 Shifted Coarsenings on page 23: 

"... there may be an opportunity to increase the matched sample size a bit
more, given the same chosen balance level, even without relaxing any
assumptions. In this situation we run the basic CEM algorithm several times,
each with a fixed value of E, and thus a fixed stratum size, but with values
of the cutpoints shifted together by different amounts. (Our software
implements this automatically.)"

This automated procedure is very interesting, because some cutpoints I am
using are somewhat arbitrary.  Is this always implimented when, e.g.,
var(#X) is used?  Is there a way to implement it with fixed breaks, e.g.,
var(1 3 5)?

Could you suggest some code for how either/both would be used?

Ben

-----Original Message-----
From: Matt Blackwell [mailto:blackwel@fas.harvard.edu] 
Sent: Tuesday, December 07, 2010 1:08 PM
To: Ben Hoen
Cc: statalist@hsphsun2.harvard.edu
Subject: Re: Seeking info on CEM cutpoints algorithms - e.g., scott, ff, fd
etc.

Hi Ben and others,

You can find a description of the different bin-cutting algorithms
along with references in section 5.5.1 of Iacus, King, and Porro
(2010), available here:

http://gking.harvard.edu/files/cem-plus.pdf

I hope that helps. We'll try to get that information into the
documentation when we update it.

Cheers,
matt.

~~~~~~~~~~~
Matthew Blackwell
PhD Candidate
Institute for Quantitative Social Science
Department of Government
Harvard University
url: http://people.fas.harvard.edu/~blackwel/

On Tue, Dec 7, 2010 at 10:42 AM, Ben Hoen <bhoen@lbl.gov> wrote:
> Statalisters (& Mathew)
>
> I have been unable to find any documentation/descriptions (via Stata help
or
> online) for the various cutpoint algorithms for CEM in Stata, namely
> "scott", "ff", "sturges" & "fd".  Can anyone direct me to more
information?
>
> Thanks in advance,
>
> Ben
>
> Ben Hoen
> bhoen@lbl.gov
>
>
>


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/