Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: combining multiple strata with single psu


From   "Stas Kolenikov" <[email protected]>
To   [email protected]
Subject   Re: st: combining multiple strata with single psu
Date   Fri, 1 Jun 2007 20:39:44 -0500

Korn & Graubard (JRSS A 1995, as well as their 1999 book) discuss
various options of dealing with the singletons; combining two
"nearest" strata, as well as combining all of the strata with single
PSUs, are "valid" options (leading to somewhat overstated standard
errors).

I am assuming you have (a pretty weird coding) with PSUs numbered
sequentially from 1 to n_h in each stratum.

list strata psu if psu==1

will give you the list of units of the first psu in each stratum, not
the list of units in singleton strata. What you need to have is first
create an indicator of the strata to be combined:

bysort strata (psu) : g byte tocombine = (psu[_N] == 1)

Then you can

egen int newpsu = group(strata psu) if tocombine

Then further

replace int newpsu = psu if ~tocombine
gen int newstrata = strata
replace newstrata = 1001 if tocombine

(or put any other reasonable stratum number in place of 1001). Hope
this is about right.

On 6/1/07, Steven Samuels <[email protected]> wrote:
On Jun 1, 2007, at 2:16 PM, <[email protected]> wrote:
> I suppose I was not quite clear enough before. I have svyset
> the data, and actually know which strata are singletons,(list
> strata psu if psu==1) the question I seem unable to solve is
> how to take those singletons, together, and make them into a
> new strata.

Combine "neighboring" strata so that you have n=2 in each new
stratum, no more if possible. (Having n=2 will facilitate using BRR
for variance estimation.) If you have an odd number of strata, the
last one will have n=3. There is a chance that the sequential stratum
numbers in your data set imply some geographic sequence  If not,
number the strata in some other way so that neighboring strata are
close by.  For geographic strata, one way is "serpentine" numbering:
123
654
789

Then neighboring sequential strata can be combined by:

gen new_strat = int(stratum/2 + .5)


You might also have other information which allows you to group
similar  strata into new ones.  Beware of grouping on the basis of
similar "outcome" variables; this will induce artificially small
within-stratum SD.

-Steve
*
*   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/


--
Stas Kolenikov
http://stas.kolenikov.name
*
*   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/



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