[Tfug] [Bulk] Re: Stallman vs Ubuntu ryryma1a

[Bexley Hall]

Hi Keith,

On 12/14/2012 5:50 PM, keith smith wrote:
> I worked for an HMO about 15 years ago.  I built a model that determined
> the risk of the groups that were utilizing under the different plans.
> This info was used by underwriting and sales and service to set pricing.
>
> All I needed was their purchase info and I could have built a predictive
> utilization model based on actuarial numbers.  If I could have obtained
> their driving info from the black boxes, now being installed in some new
> cars, I could have assessed their risk here as well.
>
> You eat too many burgers.... you became a higher risk.

What people *don't* understand is the non-obvious correlations that
"big data" can make apparent.

E.g., you figure eating lots of burgers is eating lots of red meat
which has a higher probability of resulting in heart disease, etc.
But, what you probably *wouldn't* "guess" is favoring "action"
movies is correlated with higher resting blood pressure (I'm
just making this up) which is correlated with higher probability
of heart disease.  Or, reading lots of romance novels correlates
with obesity which correlates with higher healthcare costs.  Or...

      "The ratio of the length of a man’s index finger to that of his
      ring finger may seem like a strange thing to measure, but new
      research suggests that it’s linked with penis size. The lower
      the ratio, the longer the penis."

Who the heck ever thought to correlate these two (three!) data??

> All the data is  out there today. Most of us use our credit or debit
> card for every purchase.

<grin>  A "paranoid" friend shares my refusal to use "store cards"
(Safeway, etc.) out of concern for purchase tracking, etc.  The
look of smug arrogance on his face *melted* away when I asked him
if he uses credit cards for his purchases... <grin>  He then
understood why I always had so much CASH in my wallet!

[I wonder if there are rules -- between the cc companies and the
merchants -- governing how that data can be used (by the merchant).
I.e., why would any company institute its own "store card" (NOT a
credit card, just a "reward card") if they already have access to
all this identifying information?]

> This stuff is not rocket science. Almost anyone can create one of
> these models.

And, data seldom gets stale.  Next year someone might come up
with an off-the-wall correlation that is statistically significant
(finger length vs. penis length???) that reveals something even
*more* interesting!