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

RE: st: ROC with complex data sets

From   "Taylor, Sharonda A." <>
To   <>
Subject   RE: st: ROC with complex data sets
Date   Tue, 21 Aug 2007 14:07:33 -0500

I'll give it a try. I am analyzing NHANES III data and using their final probability weight that applies to my subpopulation of interest (adolescents) per their analytic and reporting guidelines (1996 version). All of the Stata documents I have read stated that frequency and probability weights were not interchangeable. I do realize that I am new to this whole arena and truly appreciate your assistance. Thanks so much.
Sharonda Alston Taylor, MD
Adolescent Medicine Fellow
Department of Pediatrics
Section of Adolescent Medicine & Sports Medicine
Baylor College of Medicine
Texas Children's Hospital
(O) 832-822-0612 


From: on behalf of Steven Joel Hirsch Samuels
Sent: Tue 8/21/2007 1:05 PM
Subject: Re: st: ROC with complex data sets


I believe you are mistaken.

Although the weighting computations may be complicated, all 
probability surveys wind up with a "final" weight variable; call it 
"final_wt".  It is the one that would be designated the probability 
weight in -svyset- , via "[pweight=final_wt]", and it is the one you 
would round to create a frequency weight.  If that final weight 
variable is not present in your data, then contact the survey 
developers or the documentation for advice on how to calculate it.

Stratification and clustering will not affect the computation of 
sensitivity and specificity and, by extension, the ROC curves.  Only 
the p-values, standard errors, and C.I.'s will be incorrect.

Note: There can be many "final weights" in a survey data set. but 
only one will be used at a time.  In the 1999-2000 National Health 
And Nutrition Examination Survey, for example, there were exam 
weights, interview Weights, and a variety of weights for analyses of 
chemicals that were measured only in sub-samples. Problems do arise 
for investigators who want to analyse the association of two 
chemicals  not always measured together.


On Aug 21, 2007, at 10:48 AM, Taylor, Sharonda A. wrote:

> Steven,
> Thanks for the input. The problem is that I cannot simply convert 
> pweights to frequency weights  by rounding. They are derived at in 
> very different ways. Even with using svyset, I cannot run ROC 
> because it does not take into account the complex design or the 
> pweights. Any other suggestions? Am I doomed to hand calculate this?
> On Aug 20, 2007, at 8:34 AM, Taylor, Sharonda A. wrote:
>> I am very new to research. I am attempting to do ROC analysis using
>> a complex survey that uses pweights. Stata 9.2 uses frequency
>> weights. I understand that pweights and fweights are not
>> interchangeable. Are there any downloads that will allow me to run
>> the analysis with the pweights? Please advise.
> *
> *   For searches and help try:
> *
> *
> *
> The information in this e-mail may be confidential and/or 
> privileged. If you are not the intended recipient or an authorized 
> representative of the intended recipient, you are hereby notified 
> that any review, dissemination, or copying of this e-mail and its 
> attachments, if any, or the information contained herein is 
> prohibited. If you have received this e-mail in error, please 
> immediately notify the sender by return e-mail and delete this e-
> mail from your computer system.  Thank you.
> <winmail.dat>

Steven Joel Hirsch Samuels
18 Cantine's Island
Saugerties, NY 12477
Phone: 845-246-0774
EFax: 208-498-7441

*   For searches and help try:


© Copyright 1996–2022 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index