R help - Jun 2009 - classification table in logistic regression

Prof. Harrell,

My name is Armida Carbajal, I'm a graduate student intern at Sandia National
Laboratories (SNL) and am conducting some research for my thesis project at the
University of New Mexico in Statistics for SNL.

My project entails a logistic regression and I wanted to create a classification
table like the one found in SAS using the function CTABLE. I was running out of
ideas on creating code, as I am far from proficient in R. I did a google search
and found this comment.

I understand the probabilities are highly arbitrary but I thought this would be
useful. You see, I am studying a variable that is not understood in our process.
I have derived a logistic regression and well I wanted to determine how specific
and how sensitive the model was. I really didn't plan on selecting a cutoff
point. I was just going to use this piece of statistical information to give the
engineers interested in this information an idea of the sensitivity vs.
specificity of the model so that they could decide what actions to take based on
what probabilities.

I am just very curious in getting your opinion in a situation like this. Your
response would be highly appreciated.

Regards,

Armida J. Carbajal
Graduate Student Intern
Sandia National Laboratories
Dept of Analytical Technologies
ajcarba@sandia.gov
ajcarba@gmail.com
505-284-9854 (office)


***This e-mail and any attachments constitute non-public information for use
only by intended recipient (s) and may contain confidential information. If you
are not an intended recipient, please notify Sandia National Laboratories Health
Services Center Privacy Officer immediately at (505)844-7899 and permanently
delete the original of the e-mail and attachments, as well as any copies or
printouts thereof. The unauthorized use, dissemination, distribution, or
reproduction of this e-mail and any attachments is prohibited and may be
unlawful.***



	[[alternative HTML version deleted]]

Carbajal, Armida J wrote:
> Prof. Harrell,
> 
> My name is Armida Carbajal, I'm a graduate student intern at Sandia
National Laboratories (SNL) and am conducting some research for my thesis
project at the University of New Mexico in Statistics for SNL.
> 
> My project entails a logistic regression and I wanted to create a
classification table like the one found in SAS using the function CTABLE. I was
running out of ideas on creating code, as I am far from proficient in R. I did a
google search and found this comment.
> 
> I understand the probabilities are highly arbitrary but I thought this
would be useful. You see, I am studying a variable that is not understood in our
process. I have derived a logistic regression and well I wanted to determine how
specific and how sensitive the model was. I really didn't plan on selecting
a cutoff point. I was just going to use this piece of statistical information to
give the engineers interested in this information an idea of the sensitivity vs.
specificity of the model so that they could decide what actions to take based on
what probabilities.
> 
> I am just very curious in getting your opinion in a situation like this.
Your response would be highly appreciated.
> 
> Regards,
> 
> Armida J. Carbajal
> Graduate Student Intern
> Sandia National Laboratories
> Dept of Analytical Technologies
> ajcarba at sandia.gov
> ajcarba at gmail.com
> 505-284-9854 (office)
> 
>
Armida,

I regret putting CTABLE as an option on the old SAS PROC LOGIS which was 
a basis for PROC LOGISTIC.  Classification tables are arbitrary and 
misleading so I would stay away from them.

You might build a model with and without the variable of interest and 
plot the two predicted probabilities against each other for more insight 
than what is provided by a classification table.

Frank
-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University

Frank E Harrell Jr <f.harrell at vanderbilt.edu> wrote
>Armida,
>
>I regret putting CTABLE as an option on the old SAS PROC LOGIS which was 
>a basis for PROC LOGISTIC.  Classification tables are arbitrary and 
>misleading so I would stay away from them.
>
>You might build a model with and without the variable of interest and 
>plot the two predicted probabilities against each other for more insight 
>than what is provided by a classification table.
>
Frank, 

I agree with 'arbitrary' but why 'misleading'?  Purely for the
same sorts
of reasons that histograms can be misleading, or for other reasons, as well?

Plotting the probabilities is an excellent idea, I recommend it often, in a
variety
of situations.

Peter

Peter L. Flom, PhD
Statistical Consultant
www DOT peterflomconsulting DOT com