Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: noninteger frequency weights in stset


From   Steve Samuels <[email protected]>
To   [email protected]
Subject   Re: st: noninteger frequency weights in stset
Date   Mon, 19 Aug 2013 20:37:26 -0400

Paulina:

You are asking the wrong question if your data are from a complex
survey. In that case, CIs generated by -sts list- will be wrong,
whatever the weights, because the rest of the survey design is ignored.
The issue, then, is how to get the "right" CIs. Stata's standard survey
machinery can't do it because the KM estimates are not functions of a 
small number of parameters.

There may be a solution if:

(1) The data come with replicate weights (bootstrap, jackknife, BRR, or
SDR) for variance estimation.

or

(2) The data are grouped and the number of grouping intervals is small
compared to the number of observations.

A solution won't be simple in either case.

(There is actually another approach, that starts by creating your own
-bootstrap- weights with Stas Kolenikov's -bsweights- commmand
("findit"). That puts you in situation (1).)

So please look and report back.

Steve


On Aug 9, 2013, at 10:53 AM, Galezewska P.B. wrote:

Hello, 

I am having problems with including noninteger weights in "stset" command. I am using stata 12.1 for windows. 

I've been trying to apply fweight in stset command but got stata response "may not use noninteger frequency weights". I need to include weights for calculating confidence intervals for Kaplan-Meier estimates (sts list). However, if I use a multiplication factor in order to get rid of the noninteger numbers for my weights, e.g. 100 or 1000, the sample size increases and confidence intervals decrease which mess up my results. I am trying to compare the confidence intervals for estimates of two groups at a certain time (60 months). Changing the multiplication factor, changes the significance of the differences between the estimates. 
Thus, I would very appreciate if someone could explain how to use the noninteger weights in the stset command. (Stset with pweight does not give me the confidence intervals at all)


Best wishes,

Paulina Galezewska




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


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


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index