The "Process" of John Edward

[BillHoyt]

Yes, I am all ears. Always willing to learn more...

Bill? Done playing coy?

Lurker

I was hoping the realization would come to you, Lurker. Let me pose this question to shed some light on the problem:

If we can take .026 of population A and .065 of population B and declare them 60% different, then why does taking .026 of an inch and .065 of a foot and declaring them 60% different immediately trigger the error bell?


Cheers,

 

[Lurker]

I was hoping the realization would come to you, Lurker. Let me pose this question to shed some light on the problem:

If we can take .026 of population A and .065 of population B and declare them 60% different, then why does taking .026 of an inch and .065 of a foot and declaring them 60% different immediately trigger the error bell?


Cheers,

Sorry, I am a bit slow. But I consider your comparison worthless. I am examining rates. Please provide a better analogy or just speak plainly so those of us who do not have your prodigious intellect can understand you. Thank you.

Keep in mind my population B is a subset of Population A. A Sample if you will. If you want to poke holes in being unrepresentative I will agree with you. Clearly a workplace will have more of certain age groups and not represent others. Other than age, race and economic factors may skew the sample size.

Let me make an analogy so you can understand rates. Let us say the US murder rate is 5% (quantified over some period of time). In Houston let us say the muder rate is 7%. By using math we can see that (7-5)/5 = 0.4. Thus, we can say that the murder rate in Houston is 40% higher than the murder rate in the US. Note we are not adding 40% to the murder rate but are making a comparison? Capiche?

Lurker

Lurker

 

[BillHoyt]

Sorry, I am a bit slow. But I consider your comparison worthless. I am examining rates. Please provide a better analogy or just speak plainly so those of us who do not have your prodigious intellect can understand you. Thank you.

Keep in mind my population B is a subset of Population A. A Sample if you will. If you want to poke holes in being unrepresentative I will agree with you. Clearly a workplace will have more of certain age groups and not represent others. Other than age, race and economic factors may skew the sample size.

Let me make an analogy so you can understand rates. Let us say the US murder rate is 5% (quantified over some period of time). In Houston let us say the muder rate is 7%. By using math we can see that (7-5)/5 = 0.4. Thus, we can say that the murder rate in Houston is 40% higher than the murder rate in the US. Note we are not adding 40% to the murder rate but are making a comparison? Capiche?

Lurker

Lurker

Fine. Let's turn my example into rates. (And after reading this, please calm down before you post again. I set you up and you bit without even thinking. Cool down a bit before your next post) Does comparing .026 inches per hour and .065 feet per hour and declaring them 60% different make any sense to you?

Capiche?

 

[Thanz]

I invite you to read the posts carefully. Note that T'ai is proposing a particular statistical test, one of many possible. Note that I commented solely on the one-tailed chi-squared test for the low frequency initials.

Cheers,

The reason I posted is because the only objection that you had to T'ai Chi's proposed test was that it would not work with the sample size we have available. That does not tell me that T'ai Chi's test itself isn't appropriate for the work we want to do - it just tells me that we can't do it with the sample size we have now.

If I am reading Tai Chi's posts correctly, it would seem that to do his analysis one needs to have a sufficient sample size so that you would expect each letter to appear at least five times. Is that correct?

 

[BillHoyt]

Now, then: what is a "rate"? What is a "percentage"? Break them down into their component parts. Why can you not do what you did at the fundamental level?

Cheers,

 

[BillHoyt]

The reason I posted is because the only objection that you had to T'ai Chi's proposed test was that it would not work with the sample size we have available. That does not tell me that T'ai Chi's test itself isn't appropriate for the work we want to do - it just tells me that we can't do it with the sample size we have now.

If I am reading Tai Chi's posts correctly, it would seem that to do his analysis one needs to have a sufficient sample size so that you would expect each letter to appear at least five times. Is that correct?

That's the analysis he proposed for the hypothesis he proposed. That's neither my analysis nor my hypothesis. I have already demonstrated that not only can one use Poisson to test JE's "J" frequency; the result is statistically significant and (significantly) refutes my null hypothesis.

Cheers,

 

[Lurker]

Now, then: what is a "rate"? What is a "percentage"? Break them down into their component parts. Why can you not do what you did at the fundamental level?

Cheers,

Sorry, still not getting through to me. I must be very dimwitted indeed. In the context that I am using the terms, they are interchangeable.

When I see 0.046 for the letter E from the census I interpret that to mean 4.6% of the people have a name starting with E. Or a rate , 4.6 people out of a hundred would have E names if I started counting.

Sorry, Bill. I guess I am a very poor student.

Lurker

 

[BillHoyt]

Sorry, still not getting through to me. I must be very dimwitted indeed. In the context that I am using the terms, they are interchangeable.

When I see 0.046 for the letter E from the census I interpret that to mean 4.6% of the people have a name starting with E. Or a rate , 4.6 people out of a hundred would have E names if I started counting.

Sorry, Bill. I guess I am a very poor student.

Lurker

Lurker,

What is a percentage? What went wrong with saying .026 inches per hour is 60% different from .065 feet per hour?

Cheers,

 

[Kerberos]

Sorry, still not getting through to me. I must be very dimwitted indeed. In the context that I am using the terms, they are interchangeable.

When I see 0.046 for the letter E from the census I interpret that to mean 4.6% of the people have a name starting with E. Or a rate , 4.6 people out of a hundred would have E names if I started counting.

Sorry, Bill. I guess I am a very poor student.

Lurker

If it's any conciliation I don't get it either.

 

[Lurker]

Um.. your units differ? Inches versus feet? You did not divide by 12 to make the units the same. But as far as I can tell, my units are people for both calculations.

Good thing you are a bouncer in a bar. You would sure make a difficult teacher.

Lurker

 

[Lurker]

If it's any conciliation I don't get it either.

Point noted. I don't understand why Bill is dragging this out so long. He could have been quick about it and posted his answers back on the 15th. Instead, it is the 18th and I STILL don't know what I did wrong.

Instead, some 69 hours have passed and about 6 posts later and I still have nothing but cryptic posts from Bill.

Patiently waiting for the enlightenment...

Lurker

 

[BillHoyt]

Um.. your units differ? Inches versus feet? You did not divide by 12 to make the units the same. But as far as I can tell, my units are people for both calculations.

Now you're onto it. The units are different. A percentage is, like a rate, actually a numerator divided by a denominator. In doing calculations with them, one must take care to keep in mind the units being carried around.

In doing the 60% calculation, you assumed one population is a subset of another. But that is circular reasoning; you assumed an answer to the question under test. In reality, in one case, your denominator for one is "population A" and the for the other is "population B".

Cheers,

 

[Lurker]

In doing the 60% calculation, you assumed one population is a subset of another. But that is circular reasoning; you assumed an answer to the question under test. In reality, in one case, your denominator for one is "population A" and the for the other is "population B".

Sorry Bill. I STILL don't get it. Why is it I constantly see other polls, even comparisons of polls. Clearly they are SAMPLES of a larger population. By your logic, there would be no such thing as comparing SAMPLES to the population at large.

You are going to have to be MUCH clearer in what you write.

Remember, population B is a subset of population A or a SAMPLE, if you will.

Lurker

 

[Lurker]

Funny how I hear commentators say things along the line of "this is a 50% increase over the previous poll". Clearly two different polls are of two different samples, right? SAMPLE A and SAMPLE B. Or as you would term it, POPULATION A and POPULATION B.

I think you, Bill, and I are talking past each other again. I think I have plainly stated my case but am still having trouble understanding yours. Do you have a case to put forward here?

Lurker

 

[Lurker]

you assumed one population is a subset of another. But that is circular reasoning;

Um, the data provided came from the census bureau, which presumably provides the best and most complete data on names in the US.

I am comparing it to data on my company's personnel. Everyone is a US citizen here so unless they have not been counted by the census there certainly ARE a subset of the aforementioned data.

You can quibble about what they report on the census versus what their names are here but it seems pretty minor.

So, please explain how this is circular reasoning?

Gracias,

Lurker

 

[BillHoyt]

Sorry Bill. I STILL don't get it. Why is it I constantly see other polls, even comparisons of polls. Clearly they are SAMPLES of a larger population. By your logic, there would be no such thing as comparing SAMPLES to the population at large.

You are going to have to be MUCH clearer in what you write.

Remember, population B is a subset of population A or a SAMPLE, if you will.

Lurker

I'm tackling one layer at a time here. Population B is indeed a sample of A, and its representativeness is one of the base questions. That means I need to expand the "units" of the denominators. So let's try to translate those units: The census data remains population A, although we know it is not 100% accurate. This has been an issue for decades with the census and Congress has had a tizzy fit whenever the Census Bureau has suggested using statistical techniques to try to correct it. The implication here is that, even population A has a +/- to it. We know, of course, this is true of the sample of A that I had referred to as population B. This also has a +/- to it, and a larger one at that.

With that in mind, the denominators are actually "population A +/-X" and "population A +/-Y", where Y > X. We can do similar things with the standard deviation, kurtosis and skew. Any way we slice it, though, those denominators are not truly the same, and it is the sameness of those denominators that is the original question.

Cheers,

 

[Lurker]

you assumed one population is a subset of another. But that is circular reasoning; you assumed an answer to the question under test. In reality, in one case, your denominator for one is "population A" and the for the other is "population B".

Cheers,

So, provide your evidence that POP A is not a subest fo POP B. I have provided my logic for why it is. I humbly await your refutation of same.

Lurker

 

[Lurker]

I'm tackling one layer at a time here. Population B is indeed a sample of A, and its representativeness is one of the base questions. That means I need to expand the "units" of the denominators. So let's try to translate those units: The census data remains population A, although we know it is not 100% accurate. This has been an issue for decades with the census and Congress has had a tizzy fit whenever the Census Bureau has suggested using statistical techniques to try to correct it. The implication here is that, even population A has a +/- to it. We know, of course, this is true of the sample of A that I had referred to as population B. This also has a +/- to it, and a larger one at that.

With that in mind, the denominators are actually "population A +/-X" and "population A +/-Y", where Y > X. We can do similar things with the standard deviation, kurtosis and skew. Any way we slice it, though, those denominators are not truly the same, and it is the sameness of those denominators that is the original question.

I understand what you have written and to a certain degree I agree with it.

But, then no meaningful poll can ever be conducted. Your attempts to quantify the JE hits on first letters versus census data is impossible. So why are YOU advocating just such a comparison? IF the census numbers are hopelessly inaccurate and any sample (Like JE's 78) is also hopelessly inaccurate according to your definitions, what is the point? Why are you wasting time on it?

Lurker

B

 

[BillHoyt]

I understand what you have written and to a certain degree I agree with it.

But, then no meaningful poll can ever be conducted. Your attempts to quantify the JE hits on first letters versus census data is impossible. So why are YOU advocating just such a comparison? IF the census numbers are hopelessly inaccurate and any sample (Like JE's 78) is also hopelessly inaccurate according to your definitions, what is the point? Why are you wasting time on it?

Lurker

B

I'm not saying they are hopelessly inaccurate. Nor am I saying you can't compare data. I am saying you can't take the percentages of different things and compare them the way you did. (You can't take 50% of an apple, add it to 50% of an orange and get 100% of X. Unless X is a snack.) You can't simply look at 2.6% and 6.5% and say they are 60% different without knowing you are really comparing apples with apples. Percentages have units buried within them. When speaking of populations, they also have standard deviations, measurement errors, skew, and kurtosis built in. This is one of the reasons for statistics; to be able to deal with these animals given that those denominators are different.

Cheers,

 

[Lurker]

I'm not saying they are hopelessly inaccurate. Nor am I saying you can't compare data. I am saying you can't take the percentages of different things and compare them the way you did. (You can't take 50% of an apple, add it to 50% of an orange and get 100% of X. Unless X is a snack.) You can't simply look at 2.6% and 6.5% and say they are 60% different without knowing you are really comparing apples with apples. Percentages have units buried within them. When speaking of populations, they also have standard deviations, measurement errors, skew, and kurtosis built in. This is one of the reasons for statistics; to be able to deal with these animals given that those denominators are different.

Cheers,

Agreed again. But I was merely comparing averages. And clearly I was not comparing apples to oranges. I was comparing oranges from one shipment to all the known oranges. Methinks you are being unreasonably pedantic in your definitions.

I note you did not have the same problems when discussing the PASS/FAIL test for the rare/common letter test. Why do you think that is...

Lurker