home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: misc.invest
- Path: sparky!uunet!gatech!emory!sol.ctr.columbia.edu!news.columbia.edu!cucbs.chem.columbia.edu!cucbs.chem.columbia.edu!steve
- From: steve@cucbs.chem.columbia.edu (Steve Stuart)
- Subject: Re: Cyclical patterns in DJIA 1952-1980
- In-Reply-To: wendt@CS.ColoState.EDU's message of Mon, 25 Jan 1993 05:26:13 GMT
- Message-ID: <STEVE.93Jan25155959@cucbs.chem.columbia.edu>
- Sender: steve@cucbs.chem.columbia.edu (Steve Stuart)
- Organization: Center for Biomolecular Simulation, Department of Chemistry,
- Columbia University
- References: <Jan25.052613.33137@yuma.ACNS.ColoState.EDU>
- Date: Mon, 25 Jan 1993 20:59:59 GMT
- Lines: 47
-
- In article <Jan25.052613.33137@yuma.ACNS.ColoState.EDU> wendt@CS.ColoState.EDU (alan l wendt) writes:
-
- >This is the output of a program that I wrote to evaluate the average
- >return of the DJIA as related to various calendar periods. It's based
- >on the data from 1952-1980. It does not include dividends. It assumes
- >that you bought the average at the closing price the day before the
- >period began and sold at the closing price the day the period ended
- >and paid no commisions. The performance figure is the amount that
- >you would have if you started in 1952 with $1.00
-
- [ lots of data deleted ]
-
- >The following two series evaluate 9 and 10 day cycles. They
- >are intended to provide a check on the reliability of the other numbers.
- >I don't know how to do statistical tests for significant differences
- >between geometric means.
-
- [ lots more data deleted ]
-
- Interesting, but several comments jump to mind:
-
- 1) I'm not sure your 9- and 10-day cycle test for significance means
- anything. These cycles have you invested for 700 or 800 days over the
- 28-year period. The day-of-week test has you invested for 1400 days,
- and the day-of-month tests are 200 or 300 days. The means and
- variances for being invested different amounts of times should be
- different.
-
- 2) I'm no statistics whiz, but it seems to me you could do a t-test on
- the *logs* of the numbers you get. This would be a sum of the logs of
- the individual day gains, and the usual t-test formulas would hold.
- Just be careful to compare 5-day numbers to 5-day numbers and not
- 9-day numbers.
-
- 3) If you multiply the gains in any given table together, you get
- something like 3.572, which means $1.00 invested in 1952 would have
- netted you $3.57 by 1980. Does this mean the DJIA just tripled in 28
- years? That works out to be under 5% a year. This figure may very
- well be right - it just surprised me and I have nothing handy to
- check them with. (Of course dividends weren't included, and they
- might make a noticeable difference)
-
-
- --
- -Steve
-
- echo "a'rfg cnf Crpv | har" | tr '[a-m][n-z] ' '[n-z][a-m]\012' | sort | tr "\012" " " ; echo " "
-