[R] Coefficients of Logistic Regression from bootstrap - how to get them?

[Michal Figurski]

Greg and all,

Just another thought on bias and variability. As I tried to explain, I 
perceive this problem as a very practical problem.

The equation, that is the goal of this work, is supposed to serve the 
clinicians to estimate a pharmacokinetic parameter. It therefore must be 
simple and also presented in simple language, so that an average 
spreadsheet user can make use of it.

Therefore, in the end, isn't the *predictive performance* an ultimate 
measure of it all? Doesn't it account for bias and all the other stuff? 
It does tell you in how many cases you may expect to have the predicted 
value within 15% of the true value.
I apologize for my naive questions again, but aren't then the 
calculations of bias and variance, etc, just a waste of time, while you 
have it all summarized in the predictive performance?

--
Michal J. Figurski

Greg Snow wrote:
>> -----Original Message-----
>> From: r-help-bounces at r-project.org
>> [mailto:r-help-bounces at r-project.org] On Behalf Of Michal Figurski
>> Sent: Wednesday, July 23, 2008 10:22 AM
>> To: r-help at r-project.org
>> Subject: Re: [R] Coefficients of Logistic Regression from
>> bootstrap - how to get them?
>>
>> Thank you all for your words of wisdom.
>>
>> I start getting into what you mean by bootstrap. Not
>> surprisingly, it seems to be something else than I do. The
>> bootstrap is a tool, and I would rather compare it to a
>> hammer than to a gun. People say that hammer is for driving
>> nails. This situation is as if I planned to use it to break rocks.
> 
> The bootstrap is more like a whole toolbox than just a single tool.  I think part of the confusion in this discussion is because you kept asking for a hammer and Frank and others kept looking at their toolbox full of hammers and asking you which one you wanted.  Yes you can break a rock with a hammer designed to drive nails, but why not use the hammer designed to break rocks when it is easily available.
> 
>> The key point is that I don't really care about the bias or
>> variance of the mean in the model. These things are useful
>> for statisticians; regular people (like me, also a chemist)
>> do not understand them and have no use for them (well, now I
>> somewhat understand). My goal is very
>> practical: I need an equation that can predict patient's
>> outcome, based on some data, with maximum reliability and accuracy.
> 
> But to get the model with maximum reliability and accuracy you need to account for bias and minimize variability.  You may not care what those numbers are directly, but you do care indirectly about their influence on your final model.  Another instance where both sides were talking past each other.
> 
> --
> Gregory (Greg) L. Snow Ph.D.
> Statistical Data Center
> Intermountain Healthcare
> greg.snow at imail.org
> (801) 408-8111