Cold Reading Demos at TAM2

[Thanz]

And just what happens to the Poisson distribution with this alleged "overcounting"? Why didn't the "J" count move back to the mean? Thanz can't answer it. Tr'olldini can't answer it. You haven't been able to so far. Care to try?

I have explained it - and in a way, so have you. Here is what you posted earlier:

That means, that if the percentage difference between observed and expected remains the same, the significance of that observed result increases.

So, the significance increases if the percentage difference between observed and expected remains the same. We have seen this in the difference between your count and my count. Your sample and your number of J counts approximately doubled mine, with J remaining at about 21% of the total. It is as if you counted two guesses for every single real guess.

You went on to say:

If JE's repetitions of "I'm getting a J; like Joe or John" were truly random, we would expect repetitions of "I'm getting an X; like Xanadu or Xena," etc., on a random basis as well. We would expect those fluctuations to overwhelm small random perturbations in the "J"s that we see with smaller sample sizes. We would expect the percentage difference between observed and expected "J" frequencies to head to the mean; that is, to go down. That is the meaning of the fall off in the Poisson's pdf.

Here is the problem - you are not actually increasing the sample size. The underlying data in both counts is exactly the same. By overcounting the same sample, you have increased the significance of the result even though the proportions remained the same. Remember, you initially defended your overcounting on the basis that it didn't matter as everything would go up. We see that this is not true - it certainly does matter.

Your count is like taking 100 coin flips, multiplying the result by 10, and claiming that is the same as actually observing 1000 coin flips. As your explanations show, we would expect the results of 1000 coin flips to move closer to the mean than if we multiplied it out. Your counting method is like multiplying it out - which gives an artifically high count for everything, including sample size, which makes the result appear more significant.

 

[BillHoyt]

Here is the problem - you are not actually increasing the sample size.

You need remedial stat. After you achieve a passing grade there, you can join Tr'olldini in Stat 101.

 

[Thanz]

JE's choice to ejaculate a multiple or not is a Poisson process. If such ejaculation choices were Poisson with the mean matching the census data, his observed values would have turned back toward the mean.

Why is it a Poisson process? Based on what? What are you relying on for this assertion? I asked you way back in the original thread why poisson was the appropriate tool, and you never answered there either.

Stop wasting my time with this specious nonsense about letters versus names. It is patent nonsense, and you know that full well.

How can it be wasting your time when you never address it? It is not nonsense - on the contrary, your assertion that "Jane" is the same as "J" is patent nonsense. you have not made any arguments in support of your assertion - only insults to those who question it. An insult is not an argument.

Can you logically back up your position that "Jane" is the same as "J"?

 

[Thanz]

You need remedial stat. After you achieve a passing grade there, you can join Tr'olldini in Stat 101.

Ah, the refuge of Mr. Hoyt. Insults, and the laughing dog. When you have no actual arguments, I guess this is the best you can do.

 

[Thanz]

But there are three guesses: A "J"-name, "Joe" and "John".

Three.

Do you really see this as three separate guesses of J? Come on. He is guessing J, which of course opens the door wide. He then throws out a couple of common J names in the hopes that one will stick and it will be a more specific hit. But he only opens the J door once - not three times. Don't you see the difference? Any J will hit after he says J connection. John or Joe does not do anything to increase his chances of a hit - just how well that hit may be regarded by believers.

And then we haven't even taken into account the hits for "Johnny", "Joseph", "Jose", "Joshua", "Jonathan"...all within the first 100 most popular male names (US Census 1990).

Not to speak of the 16 male names within the first 100 that begin with "J".

Throwing out a "J"-name for a male is a 1-6 chance of a hit.

Yes, it does. And all of those names would hit if he had stopped at J connection. John or Joe does nothing to stop that, and certainly does not enhance it.

(We could also include the "G"-names with a "Dj"-sound. "George", "Gerald". But we won't, even though we have seen JE get a phonetic initial....)

No way to really count this, I think. We are having enough trouble agreeing when we know the letters - can you imagine the arguments if we got into "sounds like"?

Anyway: Three. 1,2,3. "J", "John", "Joe".

Here is the point I am making - his guess only gives him one shot at a hit. A good shot, to be sure, but only one shot. Saying John or Joe doesn't give him two other shots at a J hit - it just gives him a shot at a better hit within J, which is a different animal. We are not concerned with the quality of hits (or even if they are hits at all).

On the other hand, the other three readings each contain a shot at a J hit. That is unequivocally three separate chances at a hit, compared to one chance for a hit in the first reading. If we are concerned that he plays the odds to get increased chances at hits, does it really make sense to value one shot the same as three separate shots?

To put it another way, let's say that you had to bet on JE getting a hit with a J name. You are given the choice of JE saying "J, like John or Joe" to one person or JE saying "Any J name" to three separate people. You win a million dollars if he gets a hit. Which do you choose? Obviously, you choose the three separate people - as that gives you 3 shots at the million vs. one.

So in an analysis of his guesses, why would we equate the one reading with the other three combined?

 

[CFLarsen]

Thanz,

The reason why JE says "Joe" and "John" is to specify the "Dj"-sound (however you write that phonetically). That's why he opens the door more than once. There is the same sound in "George".

We have also seen that a specific relative combined with a name can easily yield a hit on the name only. Therefore, more than one "J".

 

[Thanz]

Thanz,

The reason why JE says "Joe" and "John" is to specify the "Dj"-sound (however you write that phonetically). That's why he opens the door more than once. There is the same sound in "George".

This makes no sense. We count it as more than one "J" because it might be manipulated into a name that starts with "G"? The door to a "J" hit/guess is only opened once. Whether he can manipulate it into another letter is irrelevant to a count of "J".

We have also seen that a specific relative combined with a name can easily yield a hit on the name only. Therefore, more than one "J".

I don't know what you mean by this. The sitters responses are irrelevant. If JE guesses "Grandpa Joe" and gets told it is "father in law Joe" it is still just one J guess.

 

[BillHoyt]

Ah, the refuge of Mr. Hoyt. Insults, and the laughing dog. When you have no actual arguments, I guess this is the best you can do.

Recognize this?

"Here is the problem - you are not actually increasing the sample size."

Do you recognize it as a complete contradiction of your earlier posts? As well as a completely innumerate comment? Shall I point out to you your own posts in which you entered an increased expected value into the Poisson calculator? The expected value is formed by multiplying the sample size, n, by the mean. Did the mean change? NO. So what changed? Guess that leaves the sample size, eh?

Do you still not understand what a Poisson process is? I explained it in the original thread. Do you still not understand that the multiples are also a Poisson process? Do you still not understand that the "J" frequency ought to have decreased? Apparently not as you have yet to address this.

But don't bother. I have no intention of wasting further time with you.

 

[T'ai Chi]

Fascinating. And just what happens to the Poisson distribution with this alleged "overcounting"? Why didn't the "J" count move back to the mean? Thanz can't answer it. Clancie can't answer it. How about it, Tr'olldini?

I'm not even entirely convinced that the Poisson is even applicable here.

 

[T'ai Chi]

And just what happens to the Poisson distribution with this alleged "overcounting"?

That's an easy one: You'll get smaller p-values more often, which lead to rejection of the null hypothesis more often, which reflect your natural bias more often.



Next question.

 

[T'ai Chi]

You need remedial stat. After you achieve a passing grade there, you can join Tr'olldini in Stat 101.

This from some someone who failed to answer every stat. question I put forth to him.

Be productive and go win your own Tottle award again.

 

[BillHoyt]

That's an easy one: You'll get smaller p-values more often, which lead to rejection of the null hypothesis more often, which reflect your natural bias more often.

Really? This would require the curve's shape to change, wouldn't it? What causes that shape change?

 

[Clancie]

Bill,

Another sample.

JE: I'm getting a sister with a 'B' name.

Sitter: My sister is Barbara.

JE: There's a 'D', too, a father figure.

Sitter: That's David, my dad. [2 guesses so far, right? One for D and one for B]

JE: And your grandfather had a name that started with "J"? I'm not sure which "J" name it is, but it sounds like it could be "Joe" or "John"...."Joey"..."Johnny"...."Jonathan"....All I know is that there's that strong 'J' sound at the beginning. Does that make sense to you? Did your grandfather's name start with 'J'?

And so...according to your method, the final tally from this reading is....

B: 1 guess (for sister)

D: 1 guess (for father)

J: 6 guesses (for grandfather)

Total: So, according to you, out of 8 guesses, 6 out of 8 were "J". While Thanz's method would recognize that JE guessed names for 3 people....1 of 3 (for grandpa) was a "J"

And, you are now on record as saying that your method is more accurate than his, right? That declaring out of 8 guesses, JE made six 'J' guesses is perfectly and completely accurate, a model of objective methodology?

Seriously, Bill. Are you really saying that seriously?

 

[Thanz]

Recognize this?

"Here is the problem - you are not actually increasing the sample size."

Do you recognize it as a complete contradiction of your earlier posts? As well as a completely innumerate comment? Shall I point out to you your own posts in which you entered an increased expected value into the Poisson calculator? The expected value is formed by multiplying the sample size, n, by the mean. Did the mean change? NO. So what changed? Guess that leaves the sample size, eh?

I can always count on you to misinterpret and misunderstand what I post. What is clear from the rest of the paragraph is that you have not increased the sample in real terms - only in artificial terms by your flawed count. The underlying data is exactly the same. The sample (4 transcripts from LKL) is exactly the same. You haven't gone out and counted more transcripts - that would be a real increase in sample size. You have just inflated the real sample by a factor of two, which makes the end result appear more significant than it really is. Just as if you had taken 100 coin flips and then multiplied it by 10. You haven't increased the sample in REAL terms - that is, buy actually making the extra coin flips.

Of course, this was all clear in the other post, but we can always count on you to take one sentence out of context, make some sort of attempt at ridicule, and not actually address the substantive points.

Do you still not understand what a Poisson process is? I explained it in the original thread. Do you still not understand that the multiples are also a Poisson process?

You have yet to support your bald assertion that they are, despite repeated specific requests. You are the one making the claim that Poisson is appropriate here. You have to back up that claim, if you can.

Do you still not understand that the "J" frequency ought to have decreased? Apparently not as you have yet to address this.

It ought to decrease if we actually had more of a sample - that is, more readings to go by. If we had 20 transcripts instead of only 4, and you counted 20 and I counted 4, I would agree with you.

But we counted the same 4. Your count puts an unwarranted inflationary factor into the data. That factor makes the data appear more significant than it really is.

But don't bother. I have no intention of wasting further time with you.

Run away then. You can't answer the most basic questions about the analysis you have done, and your justifications keep changing.

Remember, you originally asserted that the inflationary factor made no difference because it applied to all letters - and that it would wash out. As we have seen, this is just not so.

 

[BillHoyt]

That declaring out of 8 guesses, JE made six 'J' guesses is perfectly and completely accurate, a model of objective methodology?

What about it is not objective statistically? You keep insinuating it is somehow not objective. That would mean that the result is an artifact. Show how that statistical artifact is created. You can't. Thanz can't. Tr'olldini can't. I'm waiting.

 

[BillHoyt]

I can always count on you to misinterpret and misunderstand what I post. What is clear from the rest of the paragraph is that you have not increased the sample in real terms - only in artificial terms by your flawed count.

I perfectly understood you. You are perfectly wrong. You have artificially created your own counting laws and on that specious premise claim the count cannot be done any other way. I also perfectly understand you have no clue what a sample size is statistically.

The underlying data is exactly the same. The sample (4 transcripts from LKL) is exactly the same. You haven't gone out and counted more transcripts - that would be a real increase in sample size.

Science sometimes involves the process of returning to previously collected data with a new eye. I do not have to go collect more data to glean more information.

You have just inflated the real sample by a factor of two, which makes the end result appear more significant than it really is.

Once more with the lie, eh, Thanz. How does this happen, Thanz? The guesses are random. The multiples are also random. The "J" count should have moved to the mean. Explain this, if you can. Clancie, feel free to chime in. T'ai?

 

[T'ai Chi]

Really? This would require the curve's shape to change, wouldn't it? What causes that shape change?

No, I think it is because your observed statistic (ie. your letter J counts) are high.

 

[Thanz]

Science sometimes involves the process of returning to previously collected data with a new eye. I do not have to go collect more data to glean more information.

Does science count the same data again and again, add them together, and pretend that this is new data? That is what you have done here.

Once more with the lie, eh, Thanz. How does this happen, Thanz? The guesses are random. The multiples are also random. The "J" count should have moved to the mean. Explain this, if you can. Clancie, feel free to chime in. T'ai?

Not a lie. The truth. You have not backed up your claim that it should have moed to the mean. Your own example posted earlier that this is not so. You count 2 guesses for every one guess (on average). That is not good science.

I note that you avoid my specific questions for you to back up your claims like the plague. It does not go unnoticed. You need to back up your counting method with some real logic. Can you do it?

Which would you put your money on, if you could bet $100 on JE getting a hit: JE saying to one person, "J Connection - like John or Joe" or JE saying "J Connection" to 3 separate people at 3 separate times? Which one do you bet on? Which has a greater chance of a hit?

 

[Clancie]

Bill,

I don't want to argue the result until we've settled on the counting method. You seem to be missing the point, so let me ask it a different way.

Imagine that we are doing a "j" count and the transcript is this....

JE: I'm getting an older man, a 'J"

Sitter 1: Maybe my dad.

JE: And for you (motions to woman next to her)...there's a mother "D", like Dina, Donna, Diane, Dana...I don't know, maybe it's Dinah.

Sitter 2: No, I can't think of anyone.

JE (to Sitter 3): There's a "John", a sibling.

Sitter 3: Yes, John, my brother.

JE (back to Sitter #2): I'm still thinking of that "D" name for your mother, like "Doni, Dani, Deanna....".

Sitter #2: No, that wasn't her name.

So, Bill.... Am I right in saying that if I'm using your method, this would be 12 guesses...10 "D" guesses (even though JE made it clear he was only talking about one deceased person with a "D" name, the sitter's mother). And, 2 of the 12 guesses are "J" guesses.

Thanz's method would show "J" as being 2 out of the 3 total guesses. Your method shows "J" as only being 2 out of 12 guesses.

So...to you, the reading above would be scored as a total of 12 guesses...withl 10 "D" guesses (for one person with a "D" name) and only 2 of the 12 total guesses are "J" (though he guessed "J" twice, two different times).

That's your method and you stand by its accuracy and consistency vis a vis Thanz's.

Right?

 

[Clancie]

My point, Bill, is that JE could easily overuse "J" and prevent you from detecting it, using your method. (Though cold reading would -still- be detected using Thanz's).

For example, if JE really -is- a cold reader who overutilizes the most common initials, he could easily defeat your analysis of it by only using the initial when he over guesses the most likely letters, but to use names as well as initials when he uses less common intials.

An easy change like more frequent use of strings of names (like the following for the less common "V") would render your counting method totally useless:

JE: I'm getting a "V' name? Like Vicky, Victor, Victoria, Veronica. No? Maybe I'm getting a Vernon or Vinny.

If he did this with uncommon letters...he could go right on guessing many more "J's" (initials only, not the strings of names to go with them)...ferret out just as many "J" hits...have many more "J" guesses than "V" guesses....but, counting as you do, we would never, ever know it....

One simple change and, with your method (not Thanz's), he would have completely prevented you from detecting his cold reading patterns.