Hi Bill,
You have asked two interrelated questions: what is the meaning of statistical significance and what is the meaning of confidence. I assume from these questions that you are a layman, and so I will give you as non-technical answer as I can. When two events are correlated, and the correlation can be demonstrated statistically, we refer to this as statistically significant. In practice, we generally test the "null hypothesis" to determine a correlation. To give you a simple example,
suppose you toss a coin 100 times. You obviously expect to see 50 heads and 50 tails. Now we know intuitively that if you got 49 tails and 51 heads you would conclude that this is possible given the hypothesis that the probability of heads and tails are equal. However, as the difference between the expected value (50 in each case) and the observed value increases, we have greater basis to expect that our hypothesis is incorrect (that is, that the coin is biased). We then apply statistical tests to the data to determine whether the differences are large enough to reject the null hypothesis. The greater the confidence level, the more stringent the test.
Now the hypothesis implicit in technical analysis (regardless of the method used) is that past stock prices and volumes are predictive of future prices. So, the procedure used is to subject historical data (either individual stocks or indexes such as the DJIA or the S&P500) to such scrutiny. So far as I know, not a single serious study has been published showing any predictive value of past trading history. On the contrary, the studies I am aware of generally conclude that price movements follow a "random walk" pattern. This is not to say that no relationship exists - only that it has not been statistically demonstrated.
Alternatively we might use a methodology-specific test. For example, the following might be an interesting test. Suppose we used a computer (so as to avoid bias) to screen individual stock histories to find "head and shoulder" formations. As I understand it, technicians believe that this is a prelude to a major down-turn in prices. How often is the signal predictive? Suppose you sold short as soon as the formation was complete. How often did you make money and how often did you receive a margin call? This could easily be subjected to rigorous statistical analysis.
Before the FDA allows a drug to be marketed it is subjected clinical trials in which statistical confirmation of the efficacy of the drug is demanded. It seems prudent to me that someone risking his money using such methods would demand a similar standard of proof. Until I see such studies I will remain a non-beleiver.
Sorry for rambling, but I think this is an important issue. I hope I answered your question.
Good luck,
Paul |
