Subject: Evaluating stock-picking system
From: Alfred Vinton
Date: Sun, 12 Jan 1997 09:25:40 -0500
----------
From: Cab Vinton[SMTP:cvinton@moe.webstersite.com]
Sent: Sunday, January 12, 1997 9:22 AM
To: 'STAT-L'
Subject: Evaluating stock-picking system
After ploughing my way through an introductory stats textbook,
this is what I've come up with in terms of possible tests for
evaluating the results of a stock-picking system (returns price
deltas over N weeks for K stocks; 5 wks, 30 stocks at present):
1./ Hypergeometric distributions (or maybe binomial?), given
probabilities of deltas < or > X% for both the system and the
market as a whole.
2./ Z-tests (hypothesis testing), given mean and standard dev'n
of deltas for both the system and the market as a whole (if I
want to use a random sample from the market as a stand-in,
any guidelines on the minimum number of stocks to use?)
3./ Chi-square tests, given frequencies/ probabilities of deltas,
for both the system and the market as a whole, segregated into
bins (EG 10-15%, 15-20%, etc.)
4./ Analysis of variance tests: not sure how to do these when
working w/ a sample from a much larger population.
5./ Rank tests: not sure how or whether to apply these either . . .
Any guidance on the appropriateness of these would be greatly
appreciated:
* do the tests make assumptions that don't apply to stock prices?
& if so, does it matter?
* any suggested references in the literature (books, articles,
encyclopedia articles, etc.) of particular usefulness?
* other tests I'm not familiar with?
* other newsgroups/ mailing lists worth posting this inquiry to?
Many thanks in advance,
Cab Vinton
cvinton@moe.webstersite.com
