The bottom line question that many ask us is, what does the Torah code predict? Who is going to win the election? As explained in our Torah code tutorial, with our current methodology, there is a logical problem with using a Torah code for prediction. The logical problem is that there are many relatively compact tables that do not correspond to any event. So a single relatively compact table, even if it has an extraordinarily small p-value, cannot be used for the purpose of prediction. Saying this another way, there are many tables corresponding to no historical event that have extraordinarily small p-values. This happens because there are an extraordinarily large number of different key word combination sets that do not correspond to any historical event.
One of the two people, Obama or McCain, will be elected president of the United States. And as demonstrated, tables for both of them can be found when we look hard enough. Many of the tables we discovered have p-values that are a few out of 100. Each by itself is certainly weak evidence of any kind of encoding. It is clear that one set of tables will not correspond to the US election event. We would like to form an hypothesis about what are the characteristics of a network of associated Torah code tables that is associated with an historical event versus networks of tables that do not correspond to historical events.
Our exploration will examine the patterns of p-values in the 15 categories of key word combination sets that pertain to an event. And we will further examine the ELS commonalities in the tables. After the election event, we will form a hypothesis and use this study to develop a test statistic for the network of Torah code tables that will distinguish those tables that correspond to an event from those tables that do not correspond to an event. Once we have developed such a test statistic, we will use the developed methodology in a another historical event.
Summary of Results
Here we provide a summary showing the p-values of the best tables of the different category of experiments we did. On the previous pages, when the experiment did not result in a small p-value table we did not show the table. However, here we list the smallest estimated p-value result for each experimental category. For all estimated p-values greater than .008, the estimates are based on a 1000 trial Monte Carlo experiment. For all estimated p-values less than .008 and greater than .0002, the estimates are based on a 10,000 trial Monte Carlo experiment. For all estimated p-values less than .0002, the estimates are based on a 100,000 trial Monte Carlo experiment.
|Where When How||Topic: Who What||Num Exp||Topic: Who What||Num Exp|
|5||Where When Date||.196||12||.3895||9|
|6||Where When Year||.0085||16||.028||12|
|8||When Date Year||.065||48||.1335||36|
|9||When Date How||.0105||24||.0145||18|
|10||When Year How||.00025||32||.0125||24|
|11||Where When Date Year||.0605||48||.0405||36|
|12||Where When Date How||.272||24||.127||18|
|13||Where When Year How||.000075||32||.0255||24|
|14||When Date Year How||.015||96||.0885||72|
|15||Where When Date Year How||.0365||96||.0275||72|
|Total Num Exp||476||357|
We have analyzed an event into its who what where when and how components. We decided that the topic of the Obama or McCain event is the who and what components. This being the case, there are 15 categories of experiments. Because Obama has ELSs for the full four forms for the name while McCain has ELSs only for three of the four forms for the name, there is a difference in the total number of experiments done for each. We performed a total of 476 Obama experiments, one for each of the 476 Obama key word combination sets, and 357 McCain experiments, one for each of the 357 McCain key word combination sets.
The purpose of the analysis is to devise a test statistic using the results of the 15 different categories of the Obama and McCain experiments to distinguish between the event that will happen on November 4, 2008 from the event that will not happen on November 4, 2008. The test statistic is to be designed so that it is an appropriate one to test the Null hypothesis of no Torah code effect against the alternative that the effect is present in some significant part of the network of the 15 categories of tables produced by our event analysis. Our hope is that we can define a meaningful test statistic so that the resulting p-value of the test statistic for the experiments associated with the winner of the November 4, 2008 election will be very small and the resulting p-value for the test statistic for the experiments associated with the loser of the November 4, 2008 election will be large.
Of course, because we are using the data of these experiments to devise a test statistic, we cannot use this test statistic to make a formal test of the Null hypothesis of no Torah code effect against the alternative that the effect is present in some significant part of the network of the 15 categories of tables produced by our event analysis for the Obama-McCain election event. We can only use our analysis here to formulate a suggestion about an appropriate test statistic. We would have to use the test statistic in a setting of completely independent events to formally make the test of hypothesis of no Torah code effect against its alternative for these independent events.
What kind of characteristics will we be exploring to design the test statistic? We will be looking at redundancy, the degree to which for the same key word set, there are multiple tables whose p-values are smaller than expected under the Null hypothesis of No Torah code effect. We will be looking at within category redundancy, the degree to which there are different key word combination sets in the category which result in smaller than expected p-values. And we will be looking at the between category relationships in terms of the lattice formed by the key word combination sets of one category with respect to another. We will be exploring the tendency of key word combination sets which are super sets of other key word combination sets to have smaller p-values than those that results from those key word combination sets who are their subsets.
In 9 out of the 15 categories of experiments, the Obama results have smaller p-values. In none of the McCain categories were the p-values of the best tables smaller than .01. One of the Obama categories had a best table whose p-value was unusually small, .000075. From a quick inspection, it seems that Obama tables for the categories of 3,6,9,10,13,and 14 all have the same Obama ELS, are all located in the same text area, have related cylinder sizes, and depending on the tables, some other shared ELSs. In addition the Obama tables for category 8 and 11 have the same Obama ELS and are located in the same text area. This kind of pattern is not so strong for the McCain tables. Only the McCain tables for categories 1 and 4 are so related.
These are the kinds of network patterns we are looking for and on which we shall design a test statistic after the election results become known.