Sid writes:

> I was vague because I was looking for some general
> approaches to investigate. I am not trying to prejudge any method. But
I
> wouldn't call it vague I would call it limited. I was not looking for
> answers as to the pros and cons of performance measures and if/how
> they can or cannot be used - although any such replies are
> gratefully received - as such questions may come eventually.
>
> I don't want to spend months on just this subject because I
> see the ranking as being a parameter in a wider model. That model
> may be a one of a dozen types, I am not prejudging that either
> although I have already formulated "candidate" models - I don't see
> that the specifics of the model should affect the formulation of
> this metric, as I am going to use it as a parameter to the model - not
> as *the* target to be used for confidence intervals etc, or to
> directly estimate the likelihood of a given outcome.
>
> And more generally, I 'know' that recency weighted
> performance measures are significant as I have read papers which
> indicate this. However they have not described the techniques used
> in calculating this measure.

As I recall, your original inquiry was on how to incorporate both the
number of games and the number of days into the model. From what you're
describing now, it seems like you are gilding the lily. Any sort of
recency adjusted model ought to work well, and the simpler, the better.
There are other aspects of this project which would probably be worth
much more of your time and energy.

That's my opinion, anyway. Good luck with your project.

Steve Simon, [EMAIL PROTECTED], Standard Disclaimer.
The STATS web page has moved to
http://www.childrens-mercy.org/stats.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to