Re: [Paddlewise] swim for it?

   I have to agree that a simple attempt to do a statistical analysis using
Bayes Theorem is just not sufficient. There are too many mitigating factors
and intangibles. A good analysis could perhaps be obtained through a planned,
well-designed experiment, but of course this is not ethically and legally
possible.

Brad

Quoting Mark Perkins <marker_at_gmail.com>:

> Thank you Brad for illuminating one of my pet peeves of popularly applied
> probability.
>
> DuCharme's paper mentioned in the article is available at:
> http://article.pubs.nrc-cnrc.gc.ca/ppv/RPViewDoc?_handler_=HandleInitialGet&journal=apnm&volume=32&calyLang=eng&articleFile=h07-042.pdf
> (search for 60%)
>
> My first hope was that the journalist misinterpreted something written in
> the paper. But no, the article is true to DuCharme's exposition. My next
> hope was that DuCharme blindly copied someone else's analysis, but this
> turns out not to be so. His source is also available:
>
> http://www.redcross.ca/cmslib/general/ws_final_m2_english2006_04_19.pdf
> (Search for "VICTIM & SURVIVOR RESPONSES TO IMMERSION")
>
> As it turns out, the 2006 Drowning Report lists, for victims and surivors
> who had the choice, how many swam immediately, swam after a delay, and
> stayed with the boat. Using the raw totals, here are the relevant results:
> P{survive} = 0.53  (81 survivors, 72 victims)
> P{survive | swim immediately} = 0.51
> P{survive | swim after a delay} = 0.53
> P{survive | stay with the boat} = 0.55
>
> I haven't done the statistics, but I'm guessing that the small variation in
> these numbers with a relatively small sample is not statistically
> significant, so no claim can be made as to which course of action gives the
> best chance of survival. (Of course, the real answer is that this is
> situation dependent and must include information like, proximity to shore,
> swimming ability, PFD, clothing, etc.)
>
> While the drowning report does not report the numbers exactly as Brad
> suggested, it also does not report them as DuCharme did - so he must be held
> accountable for twisting his Bayesian inference all by himself.
>
> As for whether you should swim or stay, you'll have to decide that based on
> the conditions.
>
> -Mark
>
> PS - The good news is that of all types of boating victims, only 3% were
> kayakers, making kayaks safer than rowboats, canoes or even large powerboats
> (22%) (and yes, I'm kidding)
>
> On Nov 8, 2007 6:59 PM, Bradford R. Crain <crainb_at_pdx.edu> wrote:
>
>>   I would assume we are interested in the conditional probability of
>> survival, given that the individual swam. We would also be interested
>> in the conditional probability of survival, given the individual stayed
>> with the boat.
***************************************************************************
PaddleWise Paddling Mailing List - Any opinions or suggestions expressed
here are solely those of the writer(s). You must assume the entire
responsibility for reliance upon them. All postings copyright the author.
Submissions:     PaddleWise_at_PaddleWise.net
Subscriptions:   PaddleWise-request_at_PaddleWise.net
Website:         http://www.paddlewise.net/
***************************************************************************