[R] Firth logistic regression and odds ratio

David Winsemius dwinsemius at comcast.net
Sun Jul 24 18:03:31 CEST 2016


> On Jul 24, 2016, at 8:20 AM, Qinghua He via R-help <r-help at r-project.org> wrote:
> 
> Dear all,
> I have the following table for a breast cancer study and I was trying to calculate the odds ratio of different subtypes based on AR (AR+ as the base case).
>                   AR-    AR+Luminal A    1     19Luminal B    1     15Her2            0     4Basal-like    20    1Normal Like  2     2
> I did Firth logistic regression using the model: 
> fit <- logistf(AR ~ Age+LumA+LumB+HER2+Basal+Normal-1,data=df).summary(fit)exp(coef(fit))exp(confint(fit))


Edited from the HTML-garbled posting that appeared. Rhelp is a plain text mailing list and HTML ends up as a mess.

              coef   se(coef)  lower 0.95  upper 0.95     Chisq         p
Age     0.02675758 0.03138425 -0.03462915  0.09404934 0.7210455 0.39580118
LumA   -4.04312870 2.08208148 -8.91195072 -0.32580912 4.6216454 0.03157094
LumB   -3.74769116 1.96613514 -8.19454796 -0.16290619 4.2262682 0.03980286
HER2   -3.71796180 2.45080140 -9.50654675  0.54467296 2.8798910 0.08969209
Basal   1.12126338 1.83884710 -2.45335167  5.00553060 0.3763601 0.53955773
Normal -1.51899257 2.05597939 -5.97976670  2.40100862 0.5806537 0.44605621

> My question is: how come the p-value of Basal is so large? I was expecting a smaller p-value than LumA or LumB.

Your question is not on-topic for Rhelp., since it really has nothing to do with problems in R programming. You should be posing the question on a mailing list or a web forum where purely statistical questions are welcomed. I would expect that you would get a followup question asking you for more information, since there are many aspects of datasets that may affect p-values that you have not yet described. You for instance have not even described what "AR" might be. I'd suggest that you first read the Posting Guide and then post a question on http://stats.stackexchange.com with some more information about the study and better descriptive statistics about your population size and the covariates.


> Some extra information:
>> exp(coef(fit))

>>       Age       LumA       LumB       HER2      Basal     Normal

>> 1.02711878 0.01754250 0.02357211 0.02428341 3.06872873 0.21893233 

>> exp(confint(fit))

>>          Lower 95%   Upper 95%

>> Age    9.659636e-01   1.0986139
>> LumA   1.347687e-04   0.7219430
>> LumB   2.761551e-04   0.8496709
>> HER2   7.436339e-05   1.7240445
>> Basal  8.600484e-02 149.2362476
>> Normal 2.529416e-03  11.0343002

> If I calculate odds ratio using an online calculator based on:            AR-    AR+Basal       20    1Non-basal   4     40
> I got:
> OR:200 95% CI: 20.9510 to 1909.2138 p < 0.0001
> Can someone please help me explain?
> Thank you very much!
> Peter
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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
Alameda, CA, USA



More information about the R-help mailing list