R help - Oct 2009 - Weighted Logistic Regressions using svyglm

I?m running some logistic regressions and I?ve been trying to include weights
in the equation.  However, when I run the model, I get this warning message:

Here?s what it says:  Warning message: In eval(expr, envir, enclos) :
non-integer #successes in a binomial glm!  

I think it is because the weights are non-integer values.

What is a good way to run logistic regressions in R when using non-integer
weights?

I?ve attached the output from the R console of the two different methods
I've tried.

The first regression is unweighted.  The second regression includes the
weights in glm.  The third regression includes the weights in svyglm. 
However, despite using the same weights, I get contradicting results. 
Perhaps I am misunderstanding how to use one or both of these functions.

I'd appreciate any help you can provide.

Thanks

Brad

http://www.nabble.com/file/p25969499/Regressions%2Bwith%2BWeights.txt
Regressions+with+Weights.txt 


-- 
View this message in context:
http://www.nabble.com/Weighted-Logistic-Regressions-using-svyglm-tp25969499p25969499.html
Sent from the R help mailing list archive at Nabble.com.

Fulton wrote:
> I?m running some logistic regressions and I?ve been trying to include
weights
> in the equation.  However, when I run the model, I get this warning
message:
> 
> Here?s what it says:  Warning message: In eval(expr, envir, enclos) :
> non-integer #successes in a binomial glm!  
> 
> I think it is because the weights are non-integer values.
> 
> What is a good way to run logistic regressions in R when using non-integer
> weights?
> 
> I?ve attached the output from the R console of the two different methods
> I've tried.
> 
> The first regression is unweighted.  The second regression includes the
> weights in glm.  The third regression includes the weights in svyglm. 
> However, despite using the same weights, I get contradicting results. 
> Perhaps I am misunderstanding how to use one or both of these functions.
> 
You might be misunderstanding the use of weights in a binomial glm.
An excerpt from ?glm:

  "For a binomial GLM prior weights are used to give
   the _number of trials_ when the response is the
   _proportion of successes_"

  -Peter Ehlers
> I'd appreciate any help you can provide.
> 
> Thanks
> 
> Brad
> 
> http://www.nabble.com/file/p25969499/Regressions%2Bwith%2BWeights.txt
> Regressions+with+Weights.txt 
> 
>

On Mon, 19 Oct 2009, Fulton wrote:
>
> I?m running some logistic regressions and I?ve been trying to include
weights
> in the equation.  However, when I run the model, I get this warning
message:
>
> Here?s what it says:  Warning message: In eval(expr, envir, enclos) :
> non-integer #successes in a binomial glm!  
>
> I think it is because the weights are non-integer values.
Yes
> What is a good way to run logistic regressions in R when using non-integer
> weights?

You should use family=quasibinomial(). For svyglm() this is identical to
binomial() except that it doesn't warn about non-integer weights.

I'll update the documentation for svyglm() to make this explicit.

    -thomas

Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle