The "Process" of John Edward

[BNiles]

Now that makes sense..

I understand what is trying to be done here with J guesses exceeding the norm; however, I'm not proficient with the tools and applications to resolve the answers. I've been trying hard to follow the logic and reasoning, but I can only trust what all of you state as answers.

I feel that no matter what method or approach one uses, it is most critical that the data being considered is correct. And no disrespect, but I would have to agree with Thanz's counting method.

Don't get me wrong; I'd love for all of this to prove JE is cold reading. We could put this to bed and move on; however, when JE says, "I'm getting a J, like Joe or John." That is 1 guess. Period. Not 3. He's guessing "J" and clarifying "Joe or John".

If you said to me that you were holding an object, then asked me to guess what it was. If I said, "It's a fruit....like an apple or banana." My guess is "Fruit" not "Apple" or "Banana".

 

[Lurker]

Now that makes sense..

I feel that no matter what method or approach one uses, it is most critical that the data being considered is correct. And no disrespect, but I would have to agree with Thanz's counting method.

Count me in agreement too. As long as the non "J" letters are being counted by the same method Thanz's approach is what I believed had been done in the first place. That it was not is disappointing as I think his count more accurately defines the boundary conditions of the problem.

I think Bill sacrificed accuracy in striving to be objective.

Lurker

 

[BillHoyt]

Now that makes sense..

I understand what is trying to be done here with J guesses exceeding the norm; however, I'm not proficient with the tools and applications to resolve the answers. I've been trying hard to follow the logic and reasoning, but I can only trust what all of you state as answers.

I feel that no matter what method or approach one uses, it is most critical that the data being considered is correct. And no disrespect, but I would have to agree with Thanz's counting method.

Don't get me wrong; I'd love for all of this to prove JE is cold reading. We could put this to bed and move on; however, when JE says, "I'm getting a J, like Joe or John." That is 1 guess. Period. Not 3. He's guessing "J" and clarifying "Joe or John".

If you said to me that you were holding an object, then asked me to guess what it was. If I said, "It's a fruit....like an apple or banana." My guess is "Fruit" not "Apple" or "Banana".

Is he guessing "a person?" That is not at all clear. Ginger was a dog. Is he guessing "a first name"? That is not at all clear. He spat out "an Sh name". He got back an Sh last name and accepted that. That was a whole family! He accepted that answer! Is he guessing just one? No, we simply know that he stops at one hit. And we know that, at least on one occasion, he says one letter to identify two people.

So, what is it? One person? One dog? A whole family? In each case, JE accepted the sitter's answer as if that was his intended target.

Cheers,

 

[T'ai Chi]

, and did Bill ever answer T'ai Chi's question about tallying other letters, too?

No, he didn't.

Bill simply thinks that it is appropriate to only analyze one high frequency letter, J, instead of the other high frequency letters.

He fails to realize that, say, J could be significant, while the rest of the high frequency letters could be non significant, or a mix of significant and non significant.

Therefore, his analysis is inappropriate and misleading, and a better one, that addresses all the high frequency letters at once, is needed.

 

[BillHoyt]

No, he didn't.

Bill simply thinks that it is appropriate to only analyze one high frequency letter, J, instead of the other high frequency letters.

He fails to realize that, say, J could be significant, while the rest of the high frequency letters could be non significant, or a mix of significant and non significant.

Therefore, his analysis is inappropriate and misleading, and a better one, that addresses all the high frequency letters at once, is needed.

T'ai,

We have no idea what JE knows about letter frequencies and what he doesn't. We have no need nor any reason to assume he knows the second, third, and fourth most frequent letters in the census data. The simplest assumption is that he has learned, over time, rules of thumb about names. That, for instance, "J" is the most popular first initial.

 

[Thanz]

Do you understand that we have two different approaches to the problem, both of which have refuted the null hypothesis?

BillHoyt -

WTF is your point? It does not matter that kerberos also concluded that JE was cold reading. It makes no difference whatsoever.

Anyway, if you want to get down to the nitty-gritty, you have characterized Kerberos work as "suggestive". Further, you accepted that the control data used was not representative:

That means the control population is skewed a bit to UK names and isn't truly representative of US populations. To improve the analysis, we'd need to substitute a good US names list.

So, Kerberos test is out of whack for having the wrong control data. Your test is out of whack due to your poor counting methodology.

But all of this is irrelevant to the point I was making, and have been making for some time.

I'll make it again, on the hopes that you may eventually address it:On the test of the null hypothesis proposed by you, the only data set that rejects the null hypothesis is the data set compiled by you, using your flawed methodology.
Is that clear? I have never claimed that Kerberos did not find that JE was cold reading. We don't know what his method actually was (in terms of stat tools), and we know the control data is not what he hoped for. All of that is beside the point. If you want to get into a discussion of his numbers and method, do so with him. I don't know what he did.

 

[Thanz]

Is he guessing "a person?" That is not at all clear. Ginger was a dog. Is he guessing "a first name"? That is not at all clear. He spat out "an Sh name". He got back an Sh last name and accepted that. That was a whole family! He accepted that answer! Is he guessing just one? No, we simply know that he stops at one hit. And we know that, at least on one occasion, he says one letter to identify two people.

So, what is it? One person? One dog? A whole family? In each case, JE accepted the sitter's answer as if that was his intended target.

Cheers,

We are just counting guesses here bill. It makes no difference what the sitter says. We are counting guesses. Your argument here makes no sense. It doesn't matter if the sitter accepts a guess as a family name or a first name. What matters is that the guess is made. And unless he says otherwise, his guesses apply to first names, as that is what he commonly brings through.

 

[BillHoyt]

WTF is your point? It does not matter that kerberos also concluded that JE was cold reading. It makes no difference whatsoever.

Sure it does.

Anyway, if you want to get down to the nitty-gritty, you have characterized Kerberos work as "suggestive". Further, you accepted that the control data used was not representative:

This has no relevance to the question of whether or not the kerberos analysis refuted the null hypothesis. It did.

I'll make it again, on the hopes that you may eventually address it:On the test of the null hypothesis proposed by you, the only data set that rejects the null hypothesis is the data set compiled by you, using your flawed methodology.
Is that clear? I have never claimed that Kerberos did not find that JE was cold reading. We don't know what his method actually was (in terms of stat tools), and we know the control data is not what he hoped for. All of that is beside the point. If you want to get into a discussion of his numbers and method, do so with him. I don't know what he did.

It is clear. It is also clearly wrong. Kerberos also rejected the null hypothesis. With different data. With a different methodology.

We have two different approaches here that have rejected the null hypothesis.

 

[BillHoyt]

We are just counting guesses here bill. It makes no difference what the sitter says. We are counting guesses. Your argument here makes no sense. It doesn't matter if the sitter accepts a guess as a family name or a first name. What matters is that the guess is made. And unless he says otherwise, his guesses apply to first names, as that is what he commonly brings through.

I didn't say the sitter's answer mattered. JE's acceptance of the sitter's various answers clearly indicates that your contentions are in error. That is what matters. JE does not limit himself to one guess = one person. Sometimes he says person, but it refers to a dog. Sometimes he says one initial, and clearly says he means one person. Other times he says one initial and implicitly accepted an entire family in response.

These problems matter greatly to the counting process. You keep wanting to sweep them under the rug here, but you have not offered a solution to these problems. The solution I have proffered takes most of these problems into account.

 

[Thanz]

We have two different approaches here that have rejected the null hypothesis.

Even if you want to expand your null hypothesis now to be as broad as "cold reading", which it wasn't before, it still does not matter.

Both tests used flawed data and therefore the results cannot be trusted. Kerberos test used flawed control data. Your test used flawed counting data.

How many tests that have used correct counting data and the correct control data have rejected the null hypothesis? None.

Now, can we finally discuss what I have been trying to discuss? Are you ever going to stop wiggling around the issue, raising points about stuff I have not claimed, to address the simple point (which is not, as you claim "clearly wrong"):

On the test of the null hypothesis proposed by you, the only data set that rejects the null hypothesis is the data set compiled by you, using your flawed methodology.

 

[BillHoyt]

Even if you want to expand your null hypothesis now to be as broad as "cold reading", which it wasn't before, it still does not matter.

Both tests used flawed data and therefore the results cannot be trusted. Kerberos test used flawed control data. Your test used flawed counting data.

How many tests that have used correct counting data and the correct control data have rejected the null hypothesis? None.

Now, can we finally discuss what I have been trying to discuss? Are you ever going to stop wiggling around the issue, raising points about stuff I have not claimed, to address the simple point (which is not, as you claim "clearly wrong"):

On the test of the null hypothesis proposed by you, the only data set that rejects the null hypothesis is the data set compiled by you, using your flawed methodology.

You continue to confound the hypothesis with the experimental method. Do some homework on this before you make yourself look sillier than you already do.

 

[Clancie]

Has any discussion thread here ever resulted in Bill admitting that he was wrong, or changing his position?

Certainly none that I recall. If anything, the more wrong he appears to be, the more deeply entrenched in his position he becomes.

But...since everyone else here across the "skeptic-believer" spectrum agrees on Thanz's method, I'm going to go ahead and count the four LKL transcripts using that method and see what comes up for all the letters.

Perhaps, we can at least move the discussion forward by looking at the other letter patterns--since it is obviously futile to keep showing Bill the flaws of his method; no matter what is shown to him, it is obvious that he will not budge.

Of course, Bill is welcome to use his counting method for a tally of the rest of the letters, too.

But my prediction is that he won't, and will prefer instead to continue to argue that "J" is all that is needed.

 

[Thanz]

I didn't say the sitter's answer mattered. JE's acceptance of the sitter's various answers clearly indicates that your contentions are in error. That is what matters. JE does not limit himself to one guess = one person. Sometimes he says person, but it refers to a dog. Sometimes he says one initial, and clearly says he means one person. Other times he says one initial and implicitly accepted an entire family in response.

These problems matter greatly to the counting process. You keep wanting to sweep them under the rug here, but you have not offered a solution to these problems. The solution I have proffered takes most of these problems into account.

That is BS. Your solution CREATES problems, it doesn't solve them. It doesn't matter what JE accepts as a hit, either.

Don't you remember saying this?:

We don't care about wading through the nonsense associated with JE's hit claims.

He claims he is getting real information from real dead folk These dead folk should be calling out names that match the names of real folk, dead or alive. Therefore we should see a distribution of initials that match with initials fo real folk, dead or alive. Therefore, seeing JE call out too many of the most frequent initial says we must reject the null hypothesis that the names he calls correspond with the distribution of real names.

Look at again at the reading examples I posted. Is reading 1 REALLY equivalent to readings 2, 3, and 4 combined in terms of what we are trying to find out here? Aren't three separate and distinct guesses of a J connection different than 1 reading in which he says "Yes, a "J" - like John, or Joe?"

 

[CFLarsen]

BillHoyt, Clancie, Thanz, Lurker, whoever....

Do your own analyses and let's see what we get, OK? Time to put up or shut up. You each do your own analysis.

 

[Thanz]

You continue to confound the hypothesis with the experimental method. Do some homework on this before you make yourself look sillier than you already do.

Bill, when are you actually going to address the real issues here? Lets try taking this step by step. Let's see if you can actually address them point by point.

1.Both rejections of the null hypothesis, however you feel like defining that term, were based on flawed data of one kind or another. Kerberos on flawed control data, yours on flawed counting data.

2. The test that you propose of the null hypothesis is capable of being performed with data other than the data that you yourself have compiled.

3. If we perform your test of the null hypothesis with the raw data compiled by Kerberos (his counts of intials), the null hypothesis cannot be rejected.

4. If we perform your test of the null hypothesis with the raw data that I have compiled, we also cannot reject the null hypothesis.

Do you have anything substantive at all to say about any of these points, or do you just wish to insult me again and hope that nobody realizes that you are just runninng away?

 

[Thanz]

BillHoyt, Clancie, Thanz, Lurker, whoever....

Do your own analyses and let's see what we get, OK? Time to put up or shut up. You each do your own analysis.

Claus -

I have done this already!!!!

I went through the LKL transcripts and tallied the number of J guesses and the total number of guesses. It is on page 19 of this thread. I counted 43 guesses, of which 9 were J. If I plug this into the Poisson calculator, I get a probability of >= 9 of .128, which means that we cannot reject the null hypothesis.

 

[BillHoyt]

That is BS. Your solution CREATES problems, it doesn't solve them. It doesn't matter what JE accepts as a hit, either.

Don't you remember saying this?:

You keep claiming it creates problems and you keep basing that on the unwarranted assumption that JE is making a guess for a person. The transcripts clearly call this into question. We have dogs, we have two people for one name, we have an entire family. Although JE said, in all but one case, words that made it appear he was talking a single person, he accepted responses that made it clear that that wasn't the case.

The problem becomes believing your presuppositions or looking at the data. I chose to look at the data. We are, after all, discussing JE's process. It in no way appears to have the one person/one name relationship you presume.

And let us look at my quote in context this time, eh?

Grenard:
This is why a compendium of common and rare items, for a single sitter, is often the validating scenario rather than singling out and weighing such items as oners.Won't the probabilities for a series of facts, some common, some rare, exceed chance if they are correct and fall below chance if they are not?

Hoyt:

We don't care about wading through the nonsense associated with JE's hit claims.

I was responding to Grenard's suggestion that the better approach was to analyze claimed hits. The most charitable interpretation is that you simply misunderstood. I wouldn't want to suggest you had other motives for pulling it out of context.

 

[BillHoyt]

Bill, when are you actually going to address the real issues here? Lets try taking this step by step. Let's see if you can actually address them point by point.

1.Both rejections of the null hypothesis, however you feel like defining that term, were based on flawed data of one kind or another. Kerberos on flawed control data, yours on flawed counting data.

2. The test that you propose of the null hypothesis is capable of being performed with data other than the data that you yourself have compiled.

3. If we perform your test of the null hypothesis with the raw data compiled by Kerberos (his counts of intials), the null hypothesis cannot be rejected.

4. If we perform your test of the null hypothesis with the raw data that I have compiled, we also cannot reject the null hypothesis.

Do you have anything substantive at all to say about any of these points, or do you just wish to insult me again and hope that nobody realizes that you are just runninng away?

It is you who are running. I will discuss these when you have demonstrated competence in the fundamental concepts. Do you yet get the difference between the null hypothesis and the experimental method?

Shreiking shrilly won't help. Demonstrating an understanding will.

 

[Thanz]

You keep claiming it creates problems and you keep basing that on the unwarranted assumption that JE is making a guess for a person. The transcripts clearly call this into question. We have dogs, we have two people for one name, we have an entire family. Although JE said, in all but one case, words that made it appear he was talking a single person, he accepted responses that made it clear that that wasn't the case.

The problem becomes believing your presuppositions or looking at the data. I chose to look at the data. We are, after all, discussing JE's process. It in no way appears to have the one person/one name relationship you presume.

If we have no idea what JE is making a guess for, why count at all? Doesn't that make the whole thing a crapshoot? you say that my one-person, one guess presumption is unwarranted. Where is the evidentiary backup for your assumption that "A J - jim or Joe" refers to three persons? Use your freakin head. You say I can't assume he is making a guess for one person, or even a person at all. If that is true, it is even worse to assume that he is making 3 guesses for 3 people, which you have done.

I notice you still haven't addressed my direct points. I wonder why that is?

Also - look at the rest of your quote. It describes you doing exactly what you say I can't - attributing his guesses to letter initials of people.

 

[Thanz]

It is you who are running. I will discuss these when you have demonstrated competence in the fundamental concepts. Do you yet get the difference between the null hypothesis and the experimental method?

Shreiking shrilly won't help. Demonstrating an understanding will.

Dude, I don't think that you will ever address these. But let's take a shot, just to humour you. The hypothesis is what we are trying to prove or disprove. In this case, you are saying that this is cold reading. Despite the fact that your actual hypothesis, helpfully italicized, was "JE is cold-reading and his "J" guesses will show this by being significantly more frequent than chance alone would dictate. " If you want to truncate that to "JE is cold reading" now, whatever. If you then want to truncate your own null hypothesis to "JE is NOT cold reading", knock yourself out.

The experimental method is what we use to test the hypothesis. In this case, counting the J guesses, the total number of guesses, and using the Poisson distribution to determine significance. The Census data was used as a control, and we used it to calculate the expected number of J's in the sample.

Based on this experimental method, and based on your data, you rejected the null hypothesis. If we use your experimental method, but we use data that was collected properly either by Kerberos or myself, we cannot reject the null hypothesis.

Now, will you finally address my points?