Warp wrote:
>   So exactly where is the flaw in this math?

The flaw in the math is that you're using the same number of mexicans and 
canadians.

The percentage of interest is the percentage of illegals to legals in each 
group you want to target, *not* the absolute percentages.


You said
>   Take 100 illegal immigrants. If 90 of them are of mexican origin and
> the rest are of European origin, you certainly *can* distinguish the majority
> of them by how they look.

This is factually incorrect, *if* you have 10,000 mexicans and 30 europeans 
in the community you're policing. You will spend 10,000 investigations to 
catch 90 illegal immigrants if you investigate the ones who look mexican, 
while you'll spend 30 investigations to catch 10 illegal immigrants if you 
investigate all the europeans.

Google for Bayesian math.

>> B) there are dozens of traits more useful for distinguishing legal from 
>> illegal residents than looks.
> 
>   That may well be true.

Yet acting on that is "outrageous"?

>> Concentrate on people who look mexican?  Bad idea!
> 
>   Maybe it's a bad idea, but not because of statistics, but because of
> people's feelings and sense of privacy. That has nothing to do with math.

Yes, really, it does.  You have absolutely no idea whether genetics is the 
best distinguishing factor, yet you take an example of a different 
distinguishing factor as "outrageous".

>>>   I found it so incoherent and outlandish that I didn't know what to say.
>>> Maybe that's what you were after.
> 
>> Why did you find it outlandish?  If 90% of illegal immigrants look mexican, 
>> and 90% of illegal immigrants are christian, why is it better to focus on 
>> people who look mexican than to focus on people who look christian?
> 
>   You might as well argue "100% of illegals look like people, so why not
> concentrate on people?"

That doesn't answer the question. And yes, I wouldn't go around trying to 
find out the legal immigration status of dogs and cats.

>> The very fact that you're calling it outlandish is *exactly* the point I'm 
>> making. *That* is why focusing on people who look mexican is racist and not 
>> just statistics.
> 
>   What do you even mean by "racism"? Usually "racism" implies prejudice and
> forming preconceptions about people based on their ethnicity.

Racism: Assuming people of a particular genetic background share a common 
(usually negative) trait unrelated to their genetic background.

>   Statistical profiling based on ethnicity might be *technically* labelled
> "racism" in the sense that it segregates people based on ethnicity. However,
> if it's purely hard neutral statistics, there's no prejudiced intent. It's
> simply the result of keeping up records and calculating the statistics from
> them.

Yet, when I apply exactly the same neutral statistics to religon instead of 
genetics, it becomes outrageous and unthinkable?

>   One could argue that keeping statistics based on ethnicity is wrong, but
> curiously that happens only if the statistics make *certain* ethnicities
> look bad, while with other ethnicities it's ok. 

We're not talking about that here.

And certainly keeping statistics on ethnicity when you're talking about 
something genetic is different than when you're keeping statistics on 
ethnicity when you're talking about something entirely unrelated to genetics.

>   Statistically speaking it would be better to check only the christians
> if we want to maximize the success rate of 100 random samples.

Then why did you call it outrageous?

Note that if you're changing your mind about this, saying "I've changed my 
mind about this" is a good thing at this point.  Otherwise, I'm going to be 
confused about whether you've reconsidered and changed your mind or not. 
Because it sounds like earlier today you said "that's outrageous" and here 
you seem to be agreeing with me, so now I'm confused.

>   You are arguing that "there are too many christians and there aren't
> enough resources to check them all, hence it would be inefficient".

No, I'm arguing that if you're going to use something like that, then the 
most efficient property to use is one that most illegal immigrants all share 
and that most legal residents don't share. And "looking mexican" isn't on 
that list.

> I didn't ever claim that checking all people of a certain group would be
> *the most efficient* way. I said that it would be *more efficient* than
> purely random checks. There's a difference between "more" and "most".

Sure. But it's still far below "checking people who have done something that 
leads you to believe based on their *behavior* that they might be illegal 
aliens."

>>>   Maybe it could be a good idea to change that, then? It's not like it's
>>> something unheard of.
> 
>> Welcome to democracy.
> 
>   Democracy implies no citizenship-proving ID?

It does when the citizens consistently complain and vote against 
citizenship-proving ID. (To be clear: Of course we have citizenship-proving 
ID. We're just not required to produce it on random demand by police.)

-- 
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