Originally Posted by
dgwright99
On the contrary, it is you who are unable to grasp the significance of correlation regardless of causality.
If, in a large population, certain attributes are correalted with negative outcomes, then if I select a sub-population that excludes population members with those attributes, then axiomatically I have reduced the proprtion of negative outcomes. The only uncertainty is whether or not correlations will persist into the future.
It doesn't matter whether those attributes are causal or coincidental - whether it is a combination of conventional credit scoring factors or the number of letters in last name combined with the day of the week on which the birthday will fall in 2013.
The trap that your thinking falls into (either conciously or sub-consciously) is being overly concerned with which members of a sub-population would in reality default if no action were taken. From a risk anaysis perspective that is unknowable and irrelevant. A perfectly valid way to reduce the risk of negative outcomes within a population is to identify subpopulations the collectively exhibit higher than average negative outcomes, and to exclude that sub-population.
Beacuse the past is not a certain predictor of the future, it is not possible to be certain that excluding such a sub-population because it's past attributes were correlated with past negative outcomes will reduce the proportion of future negative outcomes, it is highly probable - provided that the past correlation has been reasonable constant over time.
I'm very familiar with what we're discussing, and I assure you based on this discussion, I'm not the one having comprehension problems. My issue is, you still haven't addressed the point here; namely, that there are both meaningful and meaningless correlations, given a certain amount of data.
The example I pointed out about a statistical survey indicating that red cars are involved in more accidents is actually quite germaine to this debate. I'm assuming you objected to it mainly because you can't get around the underlying logic.
So let's go back to that. If someone's half-brained solution is to get rid of all red cars based on this statistical correlation that they are more at risk for accidents than, say, green cars, and they do so without having access to any further data explaining whether there may simply be a higher prevalency of red cars in the total population, etc., etc., etc., then do you
really think they have actually cut their risk?
NO!
In fact, if there happen to be twice as many red cars as the next color most often involved in accidents, but the accident rate, while still higher, is anything less than 100% higher, then they've actually
increased their total relative risk exposure. Except they'll think they have reduced it, and will be patting themselves on the back for this mistake, as they won't know any better since they're relying on some half-assed and incomplete risk model. Garbage-in, garbage-out.
All Amex has available to it for this predictive behavior model based on places cardholders shop at, is their own internal data. They don't know what people are paying cash for, writing checks for, charging on VISA, or charging on Mastercard. Amex also has no idea whether someone's card use at a particular merchant has more to do with what local merchants accept which card than anything else. They also don't have any way to differentiate which guy at the liquor store (one of their known triggers) is buying azsu essencia and which guy is loading up on malt liquor.
They have nowhere near enough input data to make this system reliable, and that's readily obvious from a mile away. I guess if you really see value in this type of thing, then I don't know what to tell you, other than to point out what a 2-bit player Austin Logistics is. If their models were all they're cracked up to be, they'd have been far more successful by now. IBM originally came up with this idea back in the 1970's, and far better capitalized and far more intelligent people than Austin still couldn't make it work reliably. At least they admitted it. I think Amex has made a mistake, and what's more, it's showing up in their numbers as we speak.
But I guess at the end of the day, there is always going to be that 2% of people who will constantly argue with you and defend someone else's indefensible actions, no matter how little sense it makes. That is evidently what is happening here. You, SBM12, and Aviators99, evidently would have no problem continuing to defend Amex if they adopted a new corporate policy instructing their collections department to come to your house and kill your dog if you miss a payment. LOL. If this stuff doesn't bother you, I'm not sure what would...