[R] glm.fit: fitted probabilities numerically 0 or 1 occurred & glm.fit: algorithm did not converge

[Michael Dewey]

Your example code refers to a variable which is not in your dataset 
(repS) so I get an error message. If I assume repS is in fact rep_score 
I get another variable not found (delivery_segmentation).

I am afraid that I am unable to sort that one out so this is going to 
remain a mystery. I endorse Bert's suggestion of getting local help.

On 12/08/2016 17:24, Shivi Bhatia wrote:
> Hi Bert,
>
> Does this text file help. Apologies if this does not help as i have a
> hard time on many occasions to get a reproducible example.
>
> If this doesn't work a CSV with only 100kb of data i can share.
>
> Regards, Shivi
>
> On Fri, Aug 12, 2016 at 8:50 PM, Shivi Bhatia <shivipmp82 at gmail.com
> <mailto:shivipmp82 at gmail.com>> wrote:
>
>     Sure Burt, i will share the data after masking it.  it isn't big
>
>     regards, Shivi
>
>     On Fri, Aug 12, 2016 at 8:36 PM, Bert Gunter <bgunter.4567 at gmail.com
>     <mailto:bgunter.4567 at gmail.com>> wrote:
>
>         1. No, changing to factor will make no difference.
>
>         2. I think that most likely your problem is your model is not
>         estimable/your design matrix is singular.  You should resolve
>         this by
>         consulting with a local statistical expert or, if your data set
>         is not
>         too large or confidential, posting your full dataset using
>         dput() (see
>         ?dput for how to do this).
>
>         Cheers,
>         Bert
>         Bert Gunter
>
>         "The trouble with having an open mind is that people keep coming
>         along
>         and sticking things into it."
>         -- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
>
>
>         On Fri, Aug 12, 2016 at 7:58 AM, Shivi Bhatia
>         <shivipmp82 at gmail.com <mailto:shivipmp82 at gmail.com>> wrote:
>         > Hi Michael,
>         >
>         > There is no output as the model does not generate any
>         coefficients and
>         > simply throws this error.
>         >
>         > I hope you are not asking for a reproducible example.
>         >
>         > On Fri, Aug 12, 2016 at 7:30 PM, Michael Dewey
>         <lists at dewey.myzen.co.uk <mailto:lists at dewey.myzen.co.uk>>
>         > wrote:
>         >
>         >> Dear Shivi
>         >>
>         >> Can you show us the output?
>         >>
>         >> And please do not post in HTML as it will mangle your post into
>         >> unreadability.
>         >>
>         >> On 12/08/2016 10:10, Shivi Bhatia wrote:
>         >>
>         >>> Hi Team,
>         >>>
>         >>> I am creating *my first* Logistic regression on R Studio. I
>         am working on
>         >>> a
>         >>>
>         >>> C-SAT data where rating (score) 0-8 is a dis-sat whereas
>         9-10 are SAT. As
>         >>> these were in numeric form so i had as below created 2 classes:
>         >>>
>         >>> new$survey[new$score>=0 & new$score<=8]<- 0
>         >>> new$survey[new$score>=9]<- 1
>         >>> This works fine however the class still shows as "numeric"
>         and levels
>         >>> shows
>         >>> as "NULL". Do i still need to use "as.factor" to let R know
>         these are
>         >>> categorical variables.
>         >>>
>         >>> Also i have used the below code to run a logistic regression
>         with all the
>         >>> possible predictor variables:
>         >>> glm.fit= glm(survey ~ support_cat + region+ support_lvl+
>         skill_group+
>         >>> application_area+ functional_area+
>         >>>           repS+ case_age+ case_status+ severity_level+
>         >>>           sla_status+ delivery_segmentation, data = SFDC,
>         family =
>         >>> binomial)
>         >>>
>         >>> But it throws an error:-
>         >>> Warning messages:
>         >>> 1: glm.fit: algorithm did not converge
>         >>> 2: glm.fit: fitted probabilities numerically 0 or 1 occurred
>         >>>
>         >>> I checked online for the error and it says:
>         >>> "glm() uses an iterative re-weighted least squares
>         algorithm. The
>         >>> algorithm
>         >>> hit the maximum number of allowed iterations before signalling
>         >>> convergence.
>         >>> The default,
>         >>> documented in ?glm.control is 25."
>         >>>
>         >>> Kindly suggest on the above case and if i have to change my
>         outcome var as
>         >>> as.factor.
>         >>>
>         >>> Thank you, Shivi
>         >>>
>         >>>         [[alternative HTML version deleted]]
>         >>>
>         >>> ______________________________________________
>         >>> R-help at r-project.org <mailto:R-help at r-project.org> mailing
>         list -- To UNSUBSCRIBE and more, see
>         >>> https://stat.ethz.ch/mailman/listinfo/r-help
>         <https://stat.ethz.ch/mailman/listinfo/r-help>
>         >>> PLEASE do read the posting guide http://www.R-project.org/posti
>         >>> ng-guide.html
>         >>> and provide commented, minimal, self-contained, reproducible
>         code.
>         >>>
>         >>>
>         >> --
>         >> Michael
>         >> http://www.dewey.myzen.co.uk/home.html
>         <http://www.dewey.myzen.co.uk/home.html>
>         >>
>         >
>         >         [[alternative HTML version deleted]]
>         >
>         > ______________________________________________
>         > R-help at r-project.org <mailto:R-help at r-project.org> mailing
>         list -- To UNSUBSCRIBE and more, see
>         > https://stat.ethz.ch/mailman/listinfo/r-help
>         <https://stat.ethz.ch/mailman/listinfo/r-help>
>         > PLEASE do read the posting guide
>         http://www.R-project.org/posting-guide.html
>         <http://www.R-project.org/posting-guide.html>
>         > and provide commented, minimal, self-contained, reproducible code.
>
>
>

-- 
Michael
http://www.dewey.myzen.co.uk/home.html