[R] logistic regression

[David Winsemius]

On Apr 3, 2012, at 9:25 PM, Melrose2012 wrote:

> I am trying to plot the logistic regression of a dataset (# of  
> living flies
> vs days the flies are alive) and then fit a best-fit line to this  
> data.
>
> Here is my code:
> plot(fflies$living~fflies$day,xlab="Number of Days",ylab="Number of  
> Fruit
> Flies",main="Number of Living Fruit Flies vs Day",pch=16)
> alive <- (fflies$living)
> dead <- (fflies$living[1]-alive)
> glm.fit <- glm(cbind(alive,dead)~fflies$day,family="binomial")
> summary(glm.fit)
> lines(sort(fflies$day),fitted(glm.fit) [order (fflies$day)],col =
> "red",lwd=2)
> lines(glm.fit,col = "red",lwd=2)
>
> My problem is that, while I am pretty sure that I did the 'glm'  
> command
> correctly, when I try to plot this as a best-fit line on my data, it  
> does
> not fit (clearly on a different scale somehow).
>
> Can anyone enlighten me about what I am doing wrong?  Should I be  
> scaling
> one of these somehow?  I've tried various types of scaling but  
> nothing plots
> the line on top of the plot (no matter which I scale).

The parameter estimate is on the log(odds) scale. Two problems: A)  
Your plot is not even on the odds scale, much less the log(odds)  
scale. B) Your model assumes that the log(odds) for an event will  
increase linearly with days. However your "event" == alive seems  
reversed from what I would have expected it to be. I would have  
expected "dead" within an interval to be the event that might increase  
in probability as time increased. I also speculated that the problem  
(which was not described at all) might better fit within the framework  
of survival analysis rather than logistic regression?

-- 
David.

>
> Thanks so much for whatever help you can give me!
>
> Cheers,
> Melissa
>
> --
> View this message in context: http://r.789695.n4.nabble.com/logistic-regression-tp4530651p4530651.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

David Winsemius, MD
West Hartford, CT