The "Process" of John Edward

Posted by Bill Hoyt

If JE is cold-reading we would expect him to concentrate on those initials that are most abundant in the population. We see exactly that in this instance.
Click to expand...

Suppose he was correct 100% of the time? Do you still think the "over representation" of "J's" would still support your cold reading hypothesis if every 'J' name he got (and/or intial*) was very significant to the sitters?
 
BillHoyt said:
I might be able to help more if I understood your question. Poisson is used for events with low probability. Classic uses are events over a geographic region, events over a span of time. Things like cars arriving at an intersection in a five minute interval. Horse deaths per acre. Defects per square foot of sheet metal. In this case, names beginning with "A" or "D" or "J".
Click to expand...


I'm saying that I know the Poisson is used in some cases when you have counts. What I'm wondering, is just because we have counts, does that mean a Poisson is appropriate? That is, do we have any histogram of J counts to show that it is even remotely distributed like a Poisson? I don't doubt that it is, but I just don't want to say 'because it is a count, let's use the Poisson', when that might not be a good distributional assumption.


Close only counts in horseshoes. It means nothing. The rejection criterion is always binary: you accept or you reject. If you the significance is .049 or .019 or .0499, it still meets the criterion.
Click to expand...

Close definitely counts in statistics and the conclusions you make from it. I hope you are not claiming that (if alpha = .05) the conclusions from a p-value of .0499 are the same, because it meets the criterion as the conclusions from a p-value of .019.
 
BillHoyt said:

Yes. It adds a rejection (at the .05 level) of the hypothesis that JE's name and initial guesses match the names in the population. He overused the most frequently found forename initial, "J" significantly. If JE is cold-reading we would expect him to concentrate on those initials that are most abundant in the population. We see exactly that in this instance.
Click to expand...

What have we learned from Bill's test? I will agree with Bill that it is a good test to perform and provides some interesting results.

We see that JE guesses the letter J at a higher proportion than the average (taken from the census). From this sample, this is indisputable.

Now, we have to ask why. Bill Hoyt jumps to a conclusion that it is consistent with cold reading. Certainly a possibility and at the risk of offending Thanz I would consider it probable.

But there are other possibilities as well. Bill's conclusion is predicated on the notion that JE is familiar with the proportion of "J" or that he has learned over the years that "J" guesses seem to work more.

Perhaps JE is partial to a letter which starts his own name? Perhaps his "frames of reference" that the spirits read his mind to use have better "J" representations than certain other letters. I merely throw out other ideas to illustrate that the analysis does not indicate cold reading necessarily. It merely means he uses J more than the average. Reasons behind his overuse of J still remain a mystery.

That all being said, I will reiterate I think the reason is cold reading.

Lurker
 
Lurker,

I need a clarification.

Are you saying that JE "tunes in" (or whatever we should call it) to people who have more J's in their circle of relationships? He simply gets messages for that group more often than for others?

How is that falsifiable?
 
Tai Chi:

Your questioning whether the Poisson Distribution is valid in defining this problem is a good point.

In what I have seen there are three parameters that a distribution should have to qualify as a Poisson Distribution.

http://www.maths.unsw.edu.au/ForStudents/courses/math2899/handouts/lec4.pdf

1. N>=100

2. P<=0.01

3. NP<=20

You see, Bill, the problem is the Poisson Distribution has no upper limit. Clearly for a sample of 85 guesses 85 would be a hard limit. But the Poisson does not account for this. Further, this error is not just at the upper limit but appears in each value.

Perhaps we should be using the Binomial Distribution instead? It may not change the end results much but since p is relatively high we might want to consider it.

Just throwing my ignorant two cents into the arena.

Lurker
 
Lurker said:
Tai Chi:

Your questioning whether the Poisson Distribution is valid in defining this problem is a good point.

In what I have seen there are three parameters that a distribution should have to qualify as a Poisson Distribution.

http://www.maths.unsw.edu.au/ForStudents/courses/math2899/handouts/lec4.pdf

1. N>=100

2. P<=0.01

3. NP<=20

You see, Bill, the problem is the Poisson Distribution has no upper limit. Clearly for a sample of 85 guesses 85 would be a hard limit. But the Poisson does not account for this. Further, this error is not just at the upper limit but appears in each value.

Perhaps we should be using the Binomial Distribution instead? It may not change the end results much but since p is relatively high we might want to consider it.

Just throwing my ignorant two cents into the arena.

Lurker
Click to expand...
Lurker,

Please read that document for understanding this time. Please. This is as politely as I can put it.
 
CFLarsen said:
Lurker,

I need a clarification.

Are you saying that JE "tunes in" (or whatever we should call it) to people who have more J's in their circle of relationships? He simply gets messages for that group more often than for others?

How is that falsifiable?
Click to expand...

Glad to help clarify, I was not terribly clear in my meaning.

John Edward says the spirits read his mind and look for symbols that are in JE's frame of reference. We know he has a limited number of references. Possibly some come through better than others. for example, when he gets a "J" symbol is it a picture of him which he knows means a "J" name? Perhaps it is easier for spirits to use certain symbols in JE's lexicon. This would result in:

1. Spirits purposely choosing to communicate "J" names more often (remember it does not have to be the deceased. It can be ANYONE!) to indicate their presence.

or

2. Spirits who choose "J" symbols are more easily heard by JE.

or

3. ?

None of this is falsefiable. Just postulating possiblities. Much like cold reading is a possiblity in why he gets "J" names more often.

Lurker

P.S. And since I have your attention, I have read the Larsen List thread and generally agree with TLN. I may not always agree with your methods but you certainly ask good questions. Ones which certain believers wriggle to avoid.

I also did a quick check on some of the languages you used and from what I saw you use "possibility" when promulgating a debunking theory. It is humorous that certain believers will not even explore teh "possiblities" that you put forth. Very telling...
 
BillHoyt said:

Lurker,

Please read that document for understanding this time. Please. This is as politely as I can put it.
Click to expand...

Ah, the restrictions were for the approximation of the mean.

Still, the rest applies. Since you put for the Poisson Distribution, care to provide support that it is accurate for the problem at hand?

And why not the Binomial?

Lurker
 
Lurker,

Thanks.

I don't agree that JE has a limited number of references. It seems that he can use anything! :D
 
Lurker said:


Ah, the restrictions were for the approximation of the mean.

Still, the rest applies. Since you put for the Poisson Distribution, care to provide support that it is accurate for the problem at hand?

And why not the Binomial?

Lurker
Click to expand...
Lurker,

Your own source yields the answer. The "conditions" you cited were not to qualify a distribution as Poisson. They cite conditions under which the Poisson is a good approximation of the Binomial.

np <=20 is our condition. Poisson and Binomial in this range are almost identical. (Look at the tables in your source document.)
 
Bill:

Clearly the Poisson is best used when the probability is small, wouldn't you say?

Clearly the Poisson is best used when N is large, wouldn't you say?

Clearly this is selfevident. If you cannot see that then you truly do not understand the Poisson Distribution and I cannot help you.

Lurker
 
CFLarsen said:
Lurker,

Thanks.

I don't agree that JE has a limited number of references. It seems that he can use anything! :D
Click to expand...

Devil's advocate, my friend. Devil's advocate...

Lurker
 
Lurker said:
Bill:

Clearly the Poisson is best used when the probability is small, wouldn't you say?

Clearly the Poisson is best used when N is large, wouldn't you say?

Clearly this is selfevident. If you cannot see that then you truly do not understand the Poisson Distribution and I cannot help you.

Lurker
Click to expand...

Okay, Lurker, go back to your source. In the order of your statements, here are my responses.

Bull. Balderdash. Wrong. You need to read your source and understand it says nothing of the sort. Go back and read it. Several times if necessary. It says no such things!
 
I posted the source as an example.

Since you refuse to see the light I will go through it quite carefully so even an obstinate man can understand.

Let us say N=100
P=0.9

u=90

Use Poisson. What do you get for the probability that the count will be higher than 100? Since I am a nice guy, I will give you the answer. The answer is 1-0.8651=.1349

13.49% according to Poisson that in one hundred tries that you will get OVER 100 positive results.

Now the rest of us can see that is obviously impossible. Do you see the problem in this example?

If so, can you see that as we descend from this example to the problem we have used it for this error is diminished? The question is, how much?

Lurker
 
You see Bill, the Poisson Distribution is merely an approximation fo a more accurate distribution (binomial). So the limits in the example that you called Bull are guidelines as to whether the Poisson is accurately predicting the binomial.

You aren't under the impression that Poisson is more accurate than the Binomial, are you?

Lurker
 
Lurker,

I know nothing of statistics, but perhaps you would be kind enough to answer my question since Bill hasn't.

If JE uses "J" names and initials more often than they are expected in the general population--and, hypothetically, all of these "J" names were correct and meaningful to the sitter--does their overrepresentation in his readings still support the cold reading hypothesis?

And, if he's right about 'J', shouldn't Bill's results be consistent even if the sample increases? (I ask because there are two other LKL transcripts available on the web from 1998).

And...if the analysis of "J" is correct, can all the 26 letters be accurately calculated from the same sample, from using 78 names and initials only?
 
What? No snappy rejoinder Bill? Did my p=0.9 show you the light? Turn the light on, Bill.

Lurker
 
Lurker said:
I posted the source as an example.

Since you refuse to see the light I will go through it quite carefully so even an obstinate man can understand.

Let us say N=100
P=0.9

u=90

Use Poisson. What do you get for the probability that the count will be higher than 100? Since I am a nice guy, I will give you the answer. The answer is 1-0.8651=.1349

13.49% according to Poisson that in one hundred tries that you will get OVER 100 positive results.

Now the rest of us can see that is obviously impossible. Do you see the problem in this example?

If so, can you see that as we descend from this example to the problem we have used it for this error is diminished? The question is, how much?

Lurker
Click to expand...

ROTFLMAO! You flaming ignoramous! Holy sh**. I don't believe it. Holy sh**.

Oh, wow. You've just proven Poisson doesn't work for anything! Wow! Why did that idiot french guy invent it? BTW, fool, did you find any pairs of n and p for which this doesn't happen? What do you make of that?

Tell me where n appears in the Poisson equation. Hmm. Why not?

Holy sh**. What a stooge. You don't understand the difference between a pdf and an expectation function, do you? Holy sh**! No wonder this has gone on for pages. Holy sh**. You haven't even bothered to educate yourself. You've simply searched for bits that help support your argument. You never truly considered that you needed to learn something! Holy sh**.

Drink heavily this weekend. You're going to need it.

I'm done with you. If you don't think you need to rethink this, then be my guest. But, as far as I'm concerned, go away.
 
Clancie said:
Lurker,

Perhaps you could answer my question since Bill hasn't.
I'll give it a whirl.



If JE uses "J" names and initials more often than they are expected in the general population--and, hypothetically, all of these "J" names were correct and meaningful to the sitter--does their overrepresentation in his readings still support the cold reading hypothesis?
Click to expand...

It does not matter whether the names were meaningful to the sitter or not. Immaterial. The hypothesis that Bill has proposed is that cold readers would use popular letters more often than mere chance (in the case of "J" 13%) in order to get more hits. So, yes, the stats that he ran does lend support to his hypothesis. It does not PROVE it, merely supports it.

But, I did provide some other reasons that we may see the result that we did in the stats. I was reaching, but I am trying to show that there are other possible explanation other than cold reading. It is not an EITHER/OR scenario.


And, if valid, shouldn't Bill's results be consistent even if the sample increases? (I ask because there are two other LKL transcripts available on the web from 1998).

And (since I know nothing of statistics, I'm wondering)....if the analysis of "J" is correct, can all the 26 letters be accurately calculated from the same sample, from using 78 names and initials only?
Click to expand...

I would suggest we run the same test on the other two transcripts without using the previous data. If you can;

1. Count the "J" name guesses
2. Count the total number of guesses

then I will run the numbers and see what happens. Essentially follow the same method Bill outlined. I look forward to it.

Lurker
Click to expand...
Click to expand...
 
BillHoyt said:


Click to expand...

I guess I better go through this even slower than I thought necessary.

So Bill, how exactly did you calculate the mean for the Poisson in your "J" example? If I recall you took 0.1336*85 and arrived at 11.356. Hmm? Is that correct? Let me input the variable names for the formula you used. p*N

So, Mr Hoyt, I guess I am a stooge for following your methodology. Otherwise, please explain how you arrived at the mean for the Poisson?

Also, please give a rough idea of the limitations of the Poisson Distribution. I am starting to think you simply plug numbers into a calculator and hope for the best.

Lurker
 
You must log in or register to reply here.

Back
Top Bottom