VisorCentral.com - One Year On

Pages (24): « First ... « 3 4 5 6 7 [8] 9 10 11 12 13 » ... Last »
Show 20 posts from this thread on one page

VisorCentral.com (http://discussion.visorcentral.com/vcforum/index.php)
- Off Topic (http://discussion.visorcentral.com/vcforum/forumdisplay.php?forumid=6)
-- One Year On (http://discussion.visorcentral.com/vcforum/showthread.php?threadid=26965)

Posted by KRamsauer on 10-17-2002 06:35 PM:

quote:
Originally posted by Toby
It's not a bluff, and I'm not betting because it's a silly bet, and it also wouldn't mean anything no matter which of us won. It's like betting on a coin toss.

Then the expected payout is zero. Why wouldn't you want to take it? I'll give you $10 just for participating. There. If your hypothesis is correct, that it is simply random, you will come out ahead $10. However, I would be able to win far more than 1/3 of the bets and would therefore bleed you dry. See it isn't a coin toss because the older guy is much more likely to get heart disease.

quote:
I'm not being difficult, and I think that you understand predictive statistics less than I do if you think they'll tell you anything about a specific case from a general number.[B]I don't need the money, so I have no desire to bet. Regardless, you're still not getting the point here. It would still be 2 out of 3 against you, though, even if you did manage to do it.

It wouldn't be 2 of three against me. You dont' understand the value of statistics, though. You're saying that predicitve statistics tell you *nothing* about a particular instance, which is simply incorrect. If you were to adjust the odds in our bet to reflect the likelihood of older men getting heart disease before younger girls, then it would be a gamble because the knowledge is factored into the bet. However, since you are giving the knowledge no predictive power, you insist the odds are still 1/3 of getting it right. That is simply not the case and in the long run you will lose and lose and lose.

Posted by Toby on 10-17-2002 06:36 PM:

quote:
Originally posted by KRamsauer
And hence my mention of "most likely." I never said always. Please read what I'm writing. Your constant misunderstandings are quite frustrating.

The feeling is mutual.

quote:
This began as my stating it isn't wrong to use knowledge about someone to predict their opinions. I was being called prejudiced and hence a bad person.

You're confusing morally wrong and factually wrong.

quote:
I'm presently pointing out that of course said methods aren't perfect, but they are at use all the time and there is nothing wrong with them.

Except for the fact that there is no reliability inherent in them. At best, you're using your personal experience to make a generalization.

quote:
For instance why do advertisers advertise more geriatric products during daytime TV and more malt beverage ads during youth oriented shows? Of course not everyone watching fits their target audience but the odds are in their favor.

They've also done a bit more research and numbers than the original example at hand. Regardless, the point was that because someone is watching daytime television, you have no basis to say that they're geriatric. On average, more people may be, but one can't specify from the general. That's the point that you keep missing.

quote:
But it does guarantee that picking a point randomly from the surface of the water, odds are the depth of that water is less than 2 feet and thus odds are the person will not drown. [...]

No, it guarantees no such thing. That's the point.

Posted by Toby on 10-17-2002 06:39 PM:

quote:
Originally posted by KRamsauer
If you drown in 6 feet of water, you stand a 1/3 chance of drowning at the most. Sounds good to me, especially when you consider that a worst case scenerio. More statisticians will live than will die.

The point of the joke is that when one is crossing the river, the average depth is irrelevant if it's 6 inches deep for most of the way and 60 feet deep in the middle.

Posted by KRamsauer on 10-17-2002 06:46 PM:

quote:
Originally posted by Toby
No, it guarantees no such thing. That's the point.

But it does! Let's get technical:
You will only drown in water more than 6 feet deep.
You cannot have negative depth.
To have an average depth of 6 feet you need to have 2 points of zero depth for every point that will drown you.
There are two times more places you'll drown than you won't.
Done.

Posted by KRamsauer on 10-17-2002 06:48 PM:

quote:
Originally posted by Toby
The feeling is mutual.[B]You're confusing morally wrong and factually wrong.[B]Except for the fact that there is no reliability inherent in them. At best, you're using your personal experience to make a generalization.[B]They've also done a bit more research and numbers than the original example at hand. Regardless, the point was that because someone is watching daytime television, you have no basis to say that they're geriatric. On average, more people may be, but one can't specify from the general. That's the point that you keep missing.[B]No, it guarantees no such thing. That's the point.

AHHHHHHHHH! You keep missing the point! I'm not claiming I can tell everything. Just more than chance. When you know the population your ability to predict the properties of a sample are increased. If you doubt me, please, do us both a favor and go pick up a simple statistics book. Your maintaining that you can tell nothing about a sample from the population is so misguided as to be almost comical.

Posted by Toby on 10-17-2002 06:48 PM:

quote:
Originally posted by KRamsauer
Then the expected payout is zero. Why wouldn't you want to take it?

Have you ever done real coin-flip trials? It's quite possible to have 50 runs where 40 may come out heads and only 10 tails even though the average is 50/50 statistically.

quote:
I'll give you $10 just for participating. There. If your hypothesis is correct, that it is simply random, you will come out ahead $10. However, I would be able to win far more than 1/3 of the bets and would therefore bleed you dry. See it isn't a coin toss because the older guy is much more likely to get heart disease.

Doesn't matter. The number of trials and the selectable pool may be in either of our favors. There's no guaranteeing anything for _either_ of us.

quote:
It wouldn't be 2 of three against me.

You said that if you met a Jew, you would think they would be pro-War-on-Iraq. You met a Jew who was anti-War-on-Iraq. You thought that my wife was shorter than me. That's 2.

quote:
You dont' understand the value of statistics, though.

Yes, I do. I realize that sometimes they're jewels, and other times they're rocks. I think you _over_value statistics.

quote:
You're saying that predicitve statistics tell you *nothing* about a particular instance,

No. I'm saying that using generalized statistics (especially those which aren't really statistics and just anecdotal evidence) to try and predict a particular instance is flawed.

quote:
which is simply incorrect. If you were to adjust the odds in our bet to reflect the likelihood of older men getting heart disease before younger girls, then it would be a gamble because the knowledge is factored into the bet. However, since you are giving the knowledge no predictive power, you insist the odds are still 1/3 of getting it right. That is simply not the case and in the long run you will lose and lose and lose.

Therein lies the rub. There is no long run. In a single instance, it's no more reliable than anything else. That's the point. If you want to go around and treat people like statistics and be wrong X percentage of the time, that's your prerogative. Personally, I much prefer to just rely on only the evidence at hand rather than past anecdotal data from different samples.

Posted by Toby on 10-17-2002 06:50 PM:

quote:
Originally posted by KRamsauer
But it does! Let's get technical:
You will only drown in water more than 6 feet deep.
You cannot have negative depth.
To have an average depth of 6 feet you need to have 2 points of zero depth for every point that will drown you.
There are two times more places you'll drown than you won't.
Done.

You cannot cross the river without touching all points. You drown. Done.

Posted by KRamsauer on 10-17-2002 06:53 PM:

quote:
Originally posted by Toby

You cannot cross the river without touching all points. You drown. Done.

First the joke said nothing of crossing. I was using a single point drop as my example. Change the parameters, you change the result. Please read things carefully.

Posted by KRamsauer on 10-17-2002 06:54 PM:

Simplification

Here's what I'm saying:
"By knowing the ratio of certain traits in the population you can know the likelihood of the occurance in an individual and use that knowledge to predict with greater than chance probability the presence of that trait. Chance being 50% for the presence of a binomial trait."

Here's what you're saying:
"If you know something is present in 99% of a population, upon selecting at random a representative of that population you will only be able to predict with 50% accuracy whether or not that person has that trait."

Posted by Toby on 10-17-2002 06:57 PM:

quote:
Originally posted by KRamsauer
AHHHHHHHHH! You keep missing the point!

Consider for a moment that multiple people are telling you the same point. You are the only one whose point is seemingly being missed (in your interpretation). What are the odds that it's you?

quote:
I'm not claiming I can tell everything.

*sigh* No one thinks that you are.

quote:
Just more than chance. When you know the population your ability to predict the properties of a sample are increased.

You _don't_ know the population in this case. There's no earthly way you can. You can only base it on what _you_ know. Unless you assume that all instances are going to have the same distribution as what you know, your predictions are useless.

quote:
If you doubt me, please, do us both a favor and go pick up a simple statistics book.

I've taken graduate level statistics courses, and used to work on GIS for a living analysing Census data (which was always reviewed and never contradicted). I think I've a grasp of statistics at least at a layman's level.

quote:
Your maintaining that you can tell nothing about a sample from the population is so misguided as to be almost comical.

*sigh* Kelley, you try it.

Posted by KRamsauer on 10-17-2002 06:58 PM:

quote:
Originally posted by Toby
Yes, I do. I realize that sometimes they're jewels, and other times they're rocks. I think you _over_value statistics.

Not quite. I understand their value and their power. You are writing off the whole field because it cannot produce 100% results. Seems like a waste to me.

quote:
No. I'm saying that using generalized statistics (especially those which aren't really statistics and just anecdotal evidence) to try and predict a particular instance is flawed.

It's flawed in that it isn't perfect. It's useful in that it's better than chance. Relying on chance (in insurance, for instance) you lose while someone who relies on inference will make lots of dough.

quote:
Therein lies the rub. There is no long run. In a single instance, it's no more reliable than anything else. That's the point. If you want to go around and treat people like statistics and be wrong X percentage of the time, that's your prerogative. Personally, I much prefer to just rely on only the evidence at hand rather than past anecdotal data from different samples.

But it is more reliable than "anything else." Random selection, for instance. Dropping a ball from a building. Of course wind could carry it for miles, but odds are it'll land at the base of the building. Your refusal to recognize such relationships must cause a lot of problems, especially while driving. That car could hit you. But it's no more likely to hit you than one being driven by a drunk driver, swaying all over the road.

Posted by Toby on 10-17-2002 06:58 PM:

quote:
Originally posted by KRamsauer
First the joke said nothing of crossing. I was using a single point drop as my example. Change the parameters, you change the result. Please read things carefully.

It's implied in the joke. There are no parameters mentioned because one is supposed to use their imagination and sense of humor. Please acquire both.

Posted by KRamsauer on 10-17-2002 07:00 PM:

quote:
I've taken graduate level statistics courses, and used to work on GIS for a living analysing Census data (which was always reviewed and never contradicted). I think I've a grasp of statistics at least at a layman's level.

So you understand what I'm talking about. The whole notion of predictive validity. YOu can know something about a population and make an educated guess about a sample, though often you will be wrong, you will be more accurate than a random number generator (which is what you seem to be advocating).

Posted by Toby on 10-17-2002 07:00 PM:

Re: Simplification

quote:
Originally posted by KRamsauer
Here's what I'm saying:
"By knowing the ratio of certain traits in the population you can know the likelihood of the occurance in an individual and use that knowledge to predict with greater than chance probability the presence of that trait. Chance being 50% for the presence of a binomial trait."

Here's what you're saying:
"If you know something is present in 99% of a population, upon selecting at random a representative of that population you will only be able to predict with 50% accuracy whether or not that person has that trait."

No, here's what I'm saying:
"Just because all of the people that you've met from a certain population think a certain thing, there is no basis for generalizing that the next person from that population you meet, or that any specific number of that population think a certain thing." Keep your straw men to yourself.

Posted by KRamsauer on 10-17-2002 07:02 PM:

quote:
Originally posted by Toby
It's implied in the joke. There are no parameters mentioned because one is supposed to use their imagination and sense of humor. Please acquire both.

Sorry, in a discussion aimed at getting to the truth and realizable facts, imagination I figured was something we shouldn't be using. I've gone swimming and wading in rivers much more than I've tried to cross them so in my mind my interpreation was correct. I'm sorry I couldn't "read between the lines."

Posted by Toby on 10-17-2002 07:03 PM:

quote:
Originally posted by KRamsauer
Not quite. I understand their value and their power. You are writing off the whole field because it cannot produce 100% results. Seems like a waste to me.

Were this not getting funny, I think I'd have to scream. I never wrote off statistics. I wrote off your anecdotal evidence and experiences having any predictive validity.

Posted by KRamsauer on 10-17-2002 07:03 PM:

Re: Re: Simplification

quote:
Originally posted by Toby

No, here's what I'm saying:
"Just because all of the people that you've met from a certain population think a certain thing, there is no basis for generalizing that the next person from that population you meet, or that any specific number of that population think a certain thing." Keep your straw men to yourself.

Of course from my experiences it's hard to derive a sample large enough to derive inferential statistics. That's not what I'm claiming. I'm claiming that upon knowing the tendencies of a population (men to develop heart disease and women to get breast cancer, for instance) there is nothing wrong with using that knowledge.

Posted by Toby on 10-17-2002 07:04 PM:

quote:
Originally posted by KRamsauer
So you understand what I'm talking about. The whole notion of predictive validity. YOu can know something about a population and make an educated guess about a sample, though often you will be wrong, you will be more accurate than a random number generator (which is what you seem to be advocating).

No, you obviously are still not getting it. Look up anecdotal evidence in your statistics book.

Posted by KRamsauer on 10-17-2002 07:05 PM:

quote:
Originally posted by Toby
Were this not getting funny, I think I'd have to scream. I never wrote off statistics. I wrote off your anecdotal evidence and experiences having any predictive validity.

I never claimed my population parameters were derived from personal experience. You were writing off my using of parameters to predict samples. And doing that writes off basic principles of math and statistics.

Posted by KRamsauer on 10-17-2002 07:06 PM:

quote:
Originally posted by Toby
No, you obviously are still not getting it. Look up anecdotal evidence in your statistics book.

Anecdotes are pieces of data collected in unscientific ways (non-random sampling being the chief method). Using parameters (or inferential statistics derived with valid methods) to predict samples has nothing to do with anecdotal evidence.

All times are GMT. The time now is 04:46 AM.

Pages (24): « First ... « 3 4 5 6 7 [8] 9 10 11 12 13 » ... Last »
Show 20 posts from this thread on one page