BillHoyt said:
We're trying to see what process JE uses. We have to look at the evidence rather than make assumptions. Then we construct an hypothesis and an experimental method to test that hypothesis.
Yes, some lovely general comments from Mr. Hoyt that don't actually get to the specific issues. Fairly typical.
The problem is that you have not considered the evidence when you constructed your counting model. It makes no logical sense whatsoever, and I notice that you haven't addressed my specific points regarding this.
I'm assuming nothing. I am using the transcripts as a guide. The dog case does not fit your assumption of one person/one guess.
I assume you are referring to the dog "Ginger". I have already stated why we do not count this guess - for one, he was not guessing a name, but rather a nickname, and for another he was not guessing an initial. He was trying to make a connection based on something other than the alphabet, and therefore does not belong in our count of guesses based on the alphabet.
Neither does the single letter / two people case.
Dude, I have already stated that if JE says he is getting two people with the same intial, we count two guesses. That is not hard, now is it?
Neither does the "Sh" / whole family case.
I am not going to search the transcript for this example, but it seems that JE threw out a "Sh" guess, which was validated by the sitter as a family name. It was not thrown out by JE as a family name. If it was, we exclude it as we are studying forenames, not surnames. If he just threw out a guess WITHOUT specifying that it was a surname, we count it. It makes no difference how or even if the sitter validates the guess. All that matters is the guesses he makes. The hypothesis is that he will guess J at a significantly higher rate than could be accounted for by the general population. For that analysis, we need to accurately count the guesses. You have not done this.
Those are just three examples I found in perusing the transcript. Each of these clearly refutes the one person/one guess hypothesis. Your approach makes unwarranted assumptions that are demonstrably untrue. You need to modify your approach to account for these problems.
I find it laughable that you post these as criticisms of my method, as they apply equally to your method - if not moreso. You counted the dog guess, and the family guess just based on the number of names tossed out - without regard to the sitters response. If that is incorrect, I am sure you will let me know.
You alos, however, count "J - like Jim or John" as 3 guesses, when it is clearly one guess. So, the way I see it, your method does nothing to solve the problems you claim are in my method, and your method also has its own unique problems which completely destroy its accuracy.
Now, I have answered all of your questions. Are you ever going to address any of mine? Do you not understand that when we use a proper data collection method, your testing procedure does not reject the null hypothesis?
