Let me see if I understand the effect you're after. You want a single graph which more or less summarises overall "prophecy chatter" activity, based on the activity of the individual keywords on the page to which you linked. Is that about right? If so here's how I would go about it.
The first step is to normalise your data for all the various keywords. Some keywords have a high frequency, like "Israel" (and "New World Order", surprisingly enough), while others are quite rare, like "magog". We need to scale these data sets so that they are more or less comparable. You do this by assigning each keyword a scaling factor and dividing the raw data by this factor to produce a normalised value for that keyword.
There are three ways you could obtain this "scaling factor" for each keyword: you could use a median value, a mean value, or just choose one by hand. The first two methods require ongoing access to some historical period of data for the keyword, such as the "last 28 days" shown on the graphs to which you linked. To obtain the median value, sort the data values into a list, either in increasing or decreasing order, then select the one in the middle. (It helps to have an odd number of data values in your list, so that there is a middle, otherwise you have to take the average of the two middle values.) To obtain the mean, take the sum of all the data values and divide it by the number of data values. If you're choosing a number by hand, take a look at a graph of that keyword and pick a number representative of a "typical day". The manual approach has the disadvantage that it's not selfadjusting over time.
If it were my project, I'd elect to use the median value, but this project isn't measuring anything so scientific that the choice matters much. However you choose the "scaling factor", the process is the same: divide "today's" value for a keyword by the "scaling factor" for that keyword to get the normalised value of "today's" data. This will be a floatingpoint value of zero or more, with a typical value of close to one. The upper bound isn't limited: a value of ten would mean that today's frequency is ten times higher than what is typical of this keyword.
Optionally, you can choose a relative weight for each keyword. This is your personal assessment of how important the keyword is to measuring "prophetic chatter". In theory, you could even assign a negative weight to a value if you felt that an increase in the keyword represented a drop in the relevant chatter, but I doubt you'll want to use this trick. You simply multiply the normalised value by the relative weight to get the weighted value. If you can't be bothered, and skip this part, then you're effectively assigning a relative weight of "one" to all of them.
If you're going to choose relative weights, here's how I suggest you do it. Find a keyword which you think is of "average" significance  maybe "rapture" is a good candidate. Assign this keyword a weight of "one", and then choose weights for the other keywords based on how important you feel they are relative to the term "rapture", Maybe one "magog" is worth one and a half "raptures", in which case assign "magog" a weight of 1.5. Maybe a "globalization" is only worth half a "rapture", in which case assign "globalization" a weight of 0.5. I suggest that you also compute the average of these weights (the average weight) for scaling purposes in the next step.
Lastly, add the weighted values together, divide the result by the number of values (i.e. take the average), and divide that result by the "average weight" to obtain your index value for the day. This will have a scale such that a typical day of prophecy chatter has a value of "one" on the index, and a day with double the amount of chatter has a value of "two", and so on.
As a final note, you may want to cut down on the number of keywords you are using in your index. The keywords "Israel" and "Jerusalem" seem to have nearly identical frequencies, so there's probably not a lot of point including them both  and if you did include them both, you'd want to weight them down to about "one half" each, or else that particular pattern of keywords is going to dominate your index result. You may want to get fancy with your weightings in that regard, since weightings are (in my opinion) the tweakable aspect of such an index which really influences its usefulness.
Last edited by TFBW on Mon Nov 26, 2007 10:30 pm, edited 1 time in total.
