Hello again, Zeev. As you see, I'm now a wandering gypsy spreading my charms where'er I go. But, mainly, trying to learn something.
Thanks for the compliment but, in actual fact, it was only partly deserved. Although the 20 year model has determined that a major turn is taking place, the 16 year model has not yet forecast a major turn at p = 0.95. At the beginning of the year the upturn was only at p = 0.9 and it does not show a turn now. In fact, it shows the gold price as looking quite weak. It is because of the lack of convergence between the two models that I have not mentioned them recently. And, as you can imagine, I am also rather confused.
The models are based on multiple regression analysis which can use up to nine independent variables (eg currencies). Data is entered monthly and in one model goes back 20 years. The 16 year model starts in 1982 to avoid the 1980 bubble. The relationship between the gold price and the currencies is calculated as a multiple regression equation of the form POG(est) = a X var1 + b X var 2 ...... etc by the method of least squares. The statistical probabilities of the distribution are determined at p=0.9 and p=0.95 and the calculated "gold price" is plotted graphically at +/- p=0.95 as the upper and lower parameters of the calculated distribution, similar to the Bollinger bands. The actual gold price is plotted within these parameters.
I calculate the gold price in $US against the value of the $US and also against the currencies (valued in $US). In the $US equation, there is an unambiguous negative relationship between the POG and the value of the dollar, ie the stronger the dollar, the lower the POG. It is not simply an hypothesis, as I read someone recently remark. Of the currencies themselves, the Yen and the Swiss Franc are the most important of those having a positive correlation ie. the higher the value of the currency, the higher the gold price. Surprisingly, the Euro (= Dmark + Ffranc + Lira) has a negative correlation. As one would expect, the form of the two graphs is very similar.
Both the 20 year and 16 year model correctly showed the top turns in 1987 and 1989-90 at p=0.95. In 1993, the bottom on both was also correct at p=0.95. In 1994, the top on the 16 year was at p=0.95 and in 1996 at p=0.9 but on the 20 year the top was only at p=0.9 at both times. As I have mentioned above, the 20 year model has demonstrated a strong double bottom this year at p=0.95 but the 16 year model, so far, has only made a bottom at p=0.9 at the beginning of the year. Accordingly, because I am still awaiting a "full" bottom at p=0.95 to correspond with the 1994 top and the 1993 bottom, I am now confused as to what is going on. You will be interested that the p=0.95 level on the 16 year model corresponds to a gold price of about $245 (with the currencies as they were last month).
I trust this helps to answer some of your questions, although not directly. The models are probably too crude to respond to the actual changes which you ask about unless those changes happen at the extreme probability levels. The purpose of the models is to determine the major turns with a statistical confidence and not to track the gold price from moment to moment. One day, I would appreciate you telling me how you "grow" your turnips!
Regards
Searle
|
