[R] puzzling results from logistic regression

[Bert Gunter]

Please folks ...

On Wed, Feb 29, 2012 at 7:14 AM, Michael <comtech.usa at gmail.com> wrote:
> How did you see it's non-significant?

You need to take or review a basic statistics course. Pr(>|Z|) is your P value.

>
> Thanks!
>
> On Wed, Feb 29, 2012 at 9:10 AM, Sarah Goslee <sarah.goslee at gmail.com>wrote:
>
>> On Wed, Feb 29, 2012 at 10:02 AM, Michael <comtech.usa at gmail.com> wrote:
>> > Hi all,
>> >
>> > As you can see from below, the result is strange...
>>
>> Not really.
>>
>> > I would imagined that the bb result should be much higher and close to 1,
>> > any way to improve the fit?
>> >
>> > Any other classification methods?
>> >
>> > Thank you!
>> >
>> > data=data.frame(y=rep(c(0, 1), times=100), x=1:200)
>> > aa=glm(y~x, data=data, family=binomial(link="logit"))
>> >
>> > newdata=data.frame(x=6, y=100)
>> > bb=predict(aa, newdata=newdata, type="response")
>> > bb
>> >
>> >
>> >> bb
>> >
>> > 1
>> >
>> > 0.4929125
>>
>>
>> What did you expect? Your model is completely nonsignificant; there's no
>> way to predict y from x,
If this statement is to be parsed as "non-significant --> no
prediction", then this statement is false. The size of P values need
have nothing to do with the quality of a fitted model's predictions.

Cheers,
Bert

 and that's what your predicted value tells you.
>>
>> > summary(aa)
>>
>> Call:
>> glm(formula = y ~ x, family = binomial(link = "logit"), data = data)
>>
>> Deviance Residuals:
>>   Min      1Q  Median      3Q     Max
>> -1.190  -1.177   0.000   1.177   1.190
>>
>> Coefficients:
>>             Estimate Std. Error z value Pr(>|z|)
>> (Intercept) -0.030152   0.283924  -0.106    0.915
>> x            0.000300   0.002450   0.122    0.903
>>
>> (Dispersion parameter for binomial family taken to be 1)
>>
>>    Null deviance: 277.26  on 199  degrees of freedom
>> Residual deviance: 277.24  on 198  degrees of freedom
>> AIC: 281.24
>>
>> Number of Fisher Scoring iterations: 3
>>
>>
>> I can only assume that you didn't construct the data frame that
>> you intended to test.
>>
>> --
>> Sarah Goslee
>> http://www.functionaldiversity.org
>>
>
>        [[alternative HTML version deleted]]
>
> ______________________________________________
> 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.



-- 

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