[R] Plotting the probability curve from a logit model with 10 predictors

[Frank Harrell]

Also: require(rms); ?plot.Predict
Frank

Greg Snow wrote
> 
> Try the following:
> 
>  library(TeachingDemos)
>  ?TkPredict
>       fit.glm1 <- glm( Species=='virginica' ~ Sepal.Width+Sepal.Length,
>               data=iris, family=binomial)
>        TkPredict(fit.glm1)
> 
> (you may need to install the TeachingDemos package first if you don't
> already have it installed)
> 
> You will now see a plot that shows the predicted probability compared
> to one of the predictor variables, there are controls that you can
> then change which variable is shown on the x axis and what the value
> of the other variables are.  Play with the controls to see the effects
> of the different variables.  You can now do the same thing with other
> logistic regression models.  This also works to show nonlinear
> (polynomial, spline, etc.) fits of the variables and interactions.
> There is a button that you can click that will show the command to
> create the same plot in regular R graphics, and you can then use that
> command (and change add=TRUE to overlay multiple ones) to create a
> static plot showing the relationship.
> 
> On Fri, Jul 6, 2012 at 2:30 PM, Abraham Mathew <abmathewks@> wrote:
>> Ok, so let's say I have a logit equation outlined as Y= 2.5 + 3X1 + 2.3X2
>> +
>> 4X3 + 3.6X4 + 2.2X5
>>
>> So a one unit increase in X2 is associated with a 2.3 increase in Y,
>> regardless of what the other
>> predictor values are. So I guess instead of trying to plot of curve with
>> all the predictors accounted
>> for, I should plot each curve by itself.
>>
>> I'm still not sure how to do that with so many predictors.
>>
>> Any help would be appreciated.
>>
>>
>>
>>
>> On Thu, Jul 5, 2012 at 4:23 PM, Bert Gunter <gunter.berton@> wrote:
>>
>>> You have an about 11-D response surface, not a curve!
>>>
>>> -- Bert
>>>
>>> On Thu, Jul 5, 2012 at 2:39 PM, Abraham Mathew <abmathewks@>wrote:
>>>
>>>> I have a logit model with about 10 predictors and I am trying to plot
>>>> the
>>>> probability curve for the model.
>>>>
>>>> Y=1 = 1 / 1+e^-z  where  z=B0 + B1X1 + ... + BnXi
>>>>
>>>> If the model had only one predictor, I know to do something like below.
>>>>
>>>> mod1 = glm(factor(won) ~ as.numeric(bid), data=mydat,
>>>> family=binomial(link="logit"))
>>>>
>>>> all.x <- expand.grid(won=unique(won), bid=unique(bid))
>>>> y.hat.new <- predict(mod1, newdata=all.x, type="response")
>>>>
>>>> plot(bid<-000:250,predict(mod1,newdata=data.frame(bid<-c(000:250)),type="response"),
>>>> lwd=5, col="blue", type="l")
>>>>
>>>>
>>>> I'm not sure how to proceed when I have 10 or so predictors in the
>>>> logit
>>>> model. Do I simply expand the
>>>> expand.grid() function to include all the variables?
>>>>
>>>> So my question is how do I form a plot of a logit probability curve
>>>> when I
>>>> have 10 predictors?
>>>>
>>>> would be nice to do this in ggplot2.
>>>>
>>>> Thanks!
>>>>
>>>>
>>>> --
>>>> *Abraham Mathew
>>>> Statistical Analyst
>>>> www.amathew.com
>>>> 720-648-0108
>>>> @abmathewks*
>>>>
>>>>         [[alternative HTML version deleted]]
>>>>
>>>> ______________________________________________
>>>> R-help@ 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.
>>>>
>>>
>>>
>>>
>>> --
>>>
>>> Bert Gunter
>>> Genentech Nonclinical Biostatistics
>>>
>>> Internal Contact Info:
>>> Phone: 467-7374
>>> Website:
>>>
>>> http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm
>>>
>>>
>>>
>>
>>
>> --
>> *Abraham Mathew
>> Statistical Analyst
>> www.amathew.com
>> 720-648-0108
>> @abmathewks*
>>
>>         [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-help@ 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.
> 
> 
> 
> -- 
> Gregory (Greg) L. Snow Ph.D.
> 538280@
> 
> ______________________________________________
> R-help@ 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.
> 


-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context: http://r.789695.n4.nabble.com/Plotting-the-probability-curve-from-a-logit-model-with-10-predictors-tp4635563p4635846.html
Sent from the R help mailing list archive at Nabble.com.