Warp wrote:
>   In other words. The example I gave was correct. You then presented a
> *different* example where the same formulation doesn't work in the same
> way.

No, the example you gave was wrong.

It's like you saying "Every adult is over six feet tall."  I'm saying "No, 
that's factually true only if you pick adults for your sample set that are 
all over six feet tall."

>>  One more time:
> 
>> You can't use the math of proportions of different subgroups of illegals to 
>> target which unknowns you question. You must use different proportions of 
>> subgroups of illegals relative to subgroups of legals.  Once you do that, 
>> there's no reason to favor one particular way of splitting the total 
>> population into subgroups over another way of splitting the total population 
>> into subgroups.
> 
>   I don't see the logical line of reasoning there. It's a non-sequitur.
> 
>   "You must use different proportions of subgroups" -> "there's no reason
> to favor one particular way of splitting over another" does not logically
> follow.

You split the population of size Z into several groups, A, B, C.  Your 
suggestions for splitting criteria have most frequently been based on 
ethnicity or other properties not having to do with behavior.

Each subgroup, A, B, C, has two portions:
A' - The people in group A who are legal residents.
A'' - The people in group A who are illegal residents.
B' - The people in group B who are legal residents.
B'' - The people in group B who are illegal residents.
Etc.

The likelihood of success of stopping a person in group A and finding an 
illegal resident is not in the proportion A'' to Z, as your flawed example 
suggests. That's where I said your math is bogus.

The success is in proportion of A'' to A' (or A'' to A, depending how you 
want to measure proportions). That's the correction I applied to make it 
possible to do the statistics you wanted to do.

However, the choice of which individuals go into group A and which 
individuals go into group B and so on is most efficiently made by making A'' 
very large compared to A while making B'' very small compared to B. Then you 
can stop people in group A and efficiently find lawbreakers, while leaving 
group B alone.

The *most* efficient way to do that is to have group "A" be "the individual 
people we have actual reason to suspect are breaking the law", with group 
"B" being "the people who are doing nothing to arouse suspicion."  And this 
is a division you can't make base on the genetics, age, or any other factor 
of the person in question that the person in question cannot control.

There's no reason to *logically* believe that splitting A and B based on 
ethnicity is better or worse than splitting A and B based on religion or age 
or neighborhood of residence or value of automobile being driven.   There's 
no reason to believe and no statistics to support that anything you can tell 
simply by looking at a person has any correlation with their illegal 
immigrant status.

>   It intended to demonstrate a *point*. The numbers weren't important,
> no matter how much you like to nitpick about them.

The numbers *are* important. See above.

Now, if you're suggesting that the police should concentrate their attention 
on groups of people who demonstrate through their behavior that they might 
be illegal immigrants, nobody would be arguing with you.

>   Where? All I said is that your implication "your example doesn't work in
> this case" -> "statistics cannot be used at all" does not follow.

See above. I'll say it again here: The best statistic is probable cause.

>> Outrageous and outlandish have very similar meanings here. I simply 
>> misremembered which word you used.
> 
>   Really? I would say that outrage and strangeness are quite different
> things. Not even comparable. The first describes a feeling while the second
> describes the understandability of something.

They're not identical in meaning. But similar enough to be confusing I suppose.

>> Had you simply said "police should use statistical models to figure out 
>> which possible suspect is most likely breaking the law and concentrate on 
>> those," I would have said "They already do. It's called Probably Cause. And 
>> the current law in Arizona is targeted at preventing police from doing that."
> 
>   Yeah, using ethnicity as an example is forbidden because you are immediately
> a racist if you do.

Do you understand the difference between using someone's behavior as a 
trigger and using someone's appearance as a trigger? That using statistical 
models based on statistics that are too vague for the crime being committed 
is a bad idea? That 99% of every person you stop is going to be here 
legally, and that using statistics not based on individual behavior to 
narrow that down in any significant way isn't going to work?

It's not just ethnicity. It's anything that's irrelevant to the actual crime 
under consideration. It would be just as bad if you targeted the people in 
red automobiles, because that hasn't anything to do with the crime either.

>> Given the history of *this* country, I'm pretty sure it's in *our* best 
>> interests.
> 
>   Maybe distrust of authorities is a self-feeding process which never ends.

They have to start being trustworthy before we can start trusting them. I'm 
pretty sure it works both ways. Just like in clean cities, people put their 
trash in the trash cans in public. In cities where there's already trash all 
over the ground, people don't bother.

If we didn't get screwed by trusting our leaders, we'd be more likely to be 
trusting. But trustworthy leaders ought to be able to lead even in an 
environment of distrust.

-- 
Darren New, San Diego CA, USA (PST)
   Linux: Now bringing the quality and usability of
   open source desktop apps to your personal electronics.

Post a reply to this message