06-27-2008, 01:07 AM
I will try to explain this more clearly.
First, the so-called ???control text??? that Dave used is apparently a nearly random distribution of the Aramaic alphabet characters. This is apparent since the results for Dave???s 95 ???divine names??? fit the theoretical variance and standard deviation quite well for this control text. Hence, the control text must be pure jibberish. I think some computer-based randomization procedure was used to create this nonsence text. The fact that the theoretical variance and standard deviation values for the 95 ???devine names??? which assumes that the letters are randomly distributed fit well for the randomly distributed letters in the control text just confirms that the control text has randomly ddistributed letters, nothing more.
The logical error that Dave made was to ASSUME that all of the letters in the Peshitto text that he got with CodeFinder are "perfectly randomly" distributed. They are not! They are, in fact, organized in a very systematic fashion in the form of meaningful words, phrases, and sentences, a virtual mosaic rather than a random mess. This is a common mistake in the logical process involved in looking for "Bible Codes". Dave created a "control" text by randomly shuffling the letters in the Peshitto using a randomization routine, and the results of searching the actual Peshitto text are quite different from the control text because the Peshitto letters are not randomly distributed. No book with real words, phases, and sentences in a real language has letters that are randomly distributed. This is not a miracle. The "square root of the expected number of occurrences" can only be used to calculate 1 standard deviation if the populatation is randomly distributed because the expected number depends on random sampling. Structured equal letter spacing (ELS) sampling is not random sampling when the underlying population is not randomly distributed. Hence, the combination of non-random letters and a the structured ELS method led to biased samples.
Any real text is going to give results that are more disperse than that assumed for perfectly randomly distributed letters. Both the variances and standards deviation of the thousands of copies of each ???divine name??? that are sorted out by the ELS (equal letter spacing) CodeFinder routine are expected to be larger than the theoretical distribution for random letters. How much bigger will these varaince be? How can we estimate them?
Well, every language will present different relationships so Russian or English relationships cannot help us evaluate Aramiac relationships, To make matters even more complicated, every ELS will have different variances and standard deviations all of which are larger than the theoretical values for random letters.
In addition there is another complicating factor. You cannot ASSUME that the actual word that was chosen for the search does not affect the probability of finding the chosen combination of letters. The actual word chosen may have letters that tend to have a commonly recurrent paired relationship in the Aramaic language. Such a paired relationship can affect the result in a complex way for which the simple probability calculation does not account.
Dave added another level of variability by randomly collecting data for different ranges of ELS values and combining them in an invalid way. He thus manufactured meaninglessly large Z and so-called Chi-square values which he claimed were miraculously impossible! Through all this, remember, that about half of the ???divine names??? were found to be MORE than expected and the other about half were found to be LESS than expected. Also, typically he found about a million copies of each ???divine name??? and this so-called miracle is based on a tiny variation of about a few tenths of a percent of those observed. This is quite a stretch.
So, the only way to evaluate the true values of the variances and standard deviations for these ELS searches is to analyze the results. Firstly it is easy to show that the results fit an almost perfect normal (or Gaussian) bell shaped distribution just as expected from traditional statistical theory. Then using graphical or mathematical methods it is possible to fit these actual data to a specific normal (Gaussian) distibution. I did this for the 95 "divine name" grouped data to avoid the additional variability associated with separate ELS groups. Considering the expected differences between the results for the different ???divine names???, the results were remarkably good. The standard deviation for these data is about two times the theoretical standard deviation calculated for randomly distributed letters. There is nothing miraculous about these data!
There were no remarkable trends with about half of the observed number of "divine names" being slightly more than the calculated expected number and about half being slightly fewer. Here is a summary of the results:
(1) the distribution formed a typically Gaussian (or normal distribution) bell-shaped curve demonstrating a correction factor of about 2 times for the calculated ideal standard deviations;
(2) 47 "divine names" had fewer than the calculated "expected" number (the ideal number is half or 47.5);
(3) 48 "divine names" had more than the calculated "expected" number (the ideal number is half or 47.5);
(4) within one standard deviation of the mean there were 38 that were fewer than the mean and there were 36 that were more (the ideal number for each is 32.4);
(5) within two standard deviations of the mean there were 43 that were fewer than the mean and there were 44 that were more (the ideal number for each is 45.3);
(6) beyond two standard deviations of the mean there were 4 that were fewer than the mean and 4 that were more than the mean (the ideal number for each is 2.2).
Dave says some statistician told him that his big Z values and probabilities were correct and significantly different from the ideal theoretical values. Certainly, any statistician would say that the results he got were very significantly different from the theorercial values for randomly distributed letters and random sampling. Dave has definitely proved that the letters in the Peshitto are not randomly distributed. I certainly agree. However, this is not remarkable since we can readily understand that the letters of the Peshitto text are not randomly distributed so the sampling procedure was biased.
I would be amazed if Dave can find a qualified statistician who says that these results are a Godly miracle. Dave, if you know of one, please send me his/her e-mail address. I would like to commumicate with that person directly!
Sincerely,
Otto
First, the so-called ???control text??? that Dave used is apparently a nearly random distribution of the Aramaic alphabet characters. This is apparent since the results for Dave???s 95 ???divine names??? fit the theoretical variance and standard deviation quite well for this control text. Hence, the control text must be pure jibberish. I think some computer-based randomization procedure was used to create this nonsence text. The fact that the theoretical variance and standard deviation values for the 95 ???devine names??? which assumes that the letters are randomly distributed fit well for the randomly distributed letters in the control text just confirms that the control text has randomly ddistributed letters, nothing more.
The logical error that Dave made was to ASSUME that all of the letters in the Peshitto text that he got with CodeFinder are "perfectly randomly" distributed. They are not! They are, in fact, organized in a very systematic fashion in the form of meaningful words, phrases, and sentences, a virtual mosaic rather than a random mess. This is a common mistake in the logical process involved in looking for "Bible Codes". Dave created a "control" text by randomly shuffling the letters in the Peshitto using a randomization routine, and the results of searching the actual Peshitto text are quite different from the control text because the Peshitto letters are not randomly distributed. No book with real words, phases, and sentences in a real language has letters that are randomly distributed. This is not a miracle. The "square root of the expected number of occurrences" can only be used to calculate 1 standard deviation if the populatation is randomly distributed because the expected number depends on random sampling. Structured equal letter spacing (ELS) sampling is not random sampling when the underlying population is not randomly distributed. Hence, the combination of non-random letters and a the structured ELS method led to biased samples.
Any real text is going to give results that are more disperse than that assumed for perfectly randomly distributed letters. Both the variances and standards deviation of the thousands of copies of each ???divine name??? that are sorted out by the ELS (equal letter spacing) CodeFinder routine are expected to be larger than the theoretical distribution for random letters. How much bigger will these varaince be? How can we estimate them?
Well, every language will present different relationships so Russian or English relationships cannot help us evaluate Aramiac relationships, To make matters even more complicated, every ELS will have different variances and standard deviations all of which are larger than the theoretical values for random letters.
In addition there is another complicating factor. You cannot ASSUME that the actual word that was chosen for the search does not affect the probability of finding the chosen combination of letters. The actual word chosen may have letters that tend to have a commonly recurrent paired relationship in the Aramaic language. Such a paired relationship can affect the result in a complex way for which the simple probability calculation does not account.
Dave added another level of variability by randomly collecting data for different ranges of ELS values and combining them in an invalid way. He thus manufactured meaninglessly large Z and so-called Chi-square values which he claimed were miraculously impossible! Through all this, remember, that about half of the ???divine names??? were found to be MORE than expected and the other about half were found to be LESS than expected. Also, typically he found about a million copies of each ???divine name??? and this so-called miracle is based on a tiny variation of about a few tenths of a percent of those observed. This is quite a stretch.
So, the only way to evaluate the true values of the variances and standard deviations for these ELS searches is to analyze the results. Firstly it is easy to show that the results fit an almost perfect normal (or Gaussian) bell shaped distribution just as expected from traditional statistical theory. Then using graphical or mathematical methods it is possible to fit these actual data to a specific normal (Gaussian) distibution. I did this for the 95 "divine name" grouped data to avoid the additional variability associated with separate ELS groups. Considering the expected differences between the results for the different ???divine names???, the results were remarkably good. The standard deviation for these data is about two times the theoretical standard deviation calculated for randomly distributed letters. There is nothing miraculous about these data!
There were no remarkable trends with about half of the observed number of "divine names" being slightly more than the calculated expected number and about half being slightly fewer. Here is a summary of the results:
(1) the distribution formed a typically Gaussian (or normal distribution) bell-shaped curve demonstrating a correction factor of about 2 times for the calculated ideal standard deviations;
(2) 47 "divine names" had fewer than the calculated "expected" number (the ideal number is half or 47.5);
(3) 48 "divine names" had more than the calculated "expected" number (the ideal number is half or 47.5);
(4) within one standard deviation of the mean there were 38 that were fewer than the mean and there were 36 that were more (the ideal number for each is 32.4);
(5) within two standard deviations of the mean there were 43 that were fewer than the mean and there were 44 that were more (the ideal number for each is 45.3);
(6) beyond two standard deviations of the mean there were 4 that were fewer than the mean and 4 that were more than the mean (the ideal number for each is 2.2).
Dave says some statistician told him that his big Z values and probabilities were correct and significantly different from the ideal theoretical values. Certainly, any statistician would say that the results he got were very significantly different from the theorercial values for randomly distributed letters and random sampling. Dave has definitely proved that the letters in the Peshitto are not randomly distributed. I certainly agree. However, this is not remarkable since we can readily understand that the letters of the Peshitto text are not randomly distributed so the sampling procedure was biased.
I would be amazed if Dave can find a qualified statistician who says that these results are a Godly miracle. Dave, if you know of one, please send me his/her e-mail address. I would like to commumicate with that person directly!
Sincerely,
Otto