Re: Correlation Coefficient - was: Re: Pete Lomax
Robert Craig wrote:
>
> don cole wrote:
> > Could be a little [OT] but would those two streams of data be the day to
> > day closing price of that stock, mutual fund, currency exchange rate or
> > whatever?
> > I never thought of doing that. (Keeping track of all that data).
> > I've just been looking at the day to day results.
>
> I've been tracking the weekly closing unit price
> on a bunch of mutual funds offered by one company.
> I can easily copy/paste the data from their Web page.
>
> You could track the daily, weekly, monthly, whatever,
> closing prices of various individual stocks or anything else you like.
> A lot of that data is readily available on the Web.
>
> You can correlate *any* two things. Daily temperatures versus sales of
> air conditioners the following week. Sun spot activity versus the
> Dow Jones average. You might even correlate a time series against itself,
> but delayed by one or more time periods. e.g. if the stock market
> goes up one week, is it more likely to go up again the next week, or down?
> A correlation of the market index against itself, but shifted
> by one week might give you a clue. (I recall I found a slight
> tendency for the Toronto market to reverse itself, i.e. a small
> negative correlation between the series, and the series shifted by a week.)
>
> I'm assuming that if two things are correlated day-to-day or
> week-to-week they will be similarly correlated over the long haul.
Not quite true, because the regression line may change over the time.
There may still be a correlation, but possibly a non linear or piecewise linear
one. Standard correlation coefficient is basically linear and will give you a
wrongly weak value, since it measures the consistency of data fitting to an
y=ax+b pattern.
CChris
> I also compute the standard deviation of each series to get an idea
> of how "volatile" each fund is. Investing in foreign markets is
> good as far as correlations are concerned, but then you have
> added risk from currency fluctuations. Currency fluctuations are
> basically risk without reward, since you can't assume that
> one currency will rise against the other over a long period of time.
> When the U.S. markets rise, I might still lose because of the
> exchange rate between U.S. and Canadian dollars. The Canadian dollar
> has risen sharply (62 cents U.S. about 5 years ago. 96 cents now).
> That makes most U.S. investments over the past few years look bad
> to Canadians, who have to buy their beer and most other goods
> and services with Canadian dollars. On the other hand, Canadian
> stock funds over the past 5 years must have been an excellent
> investment for an American. (Doubling * 1.5 for currency = tripling in value).
>
> Regards,
> Rob Craig
> Rapid Deployment Software
> <a href="http://www.RapidEuphoria.com">http://www.RapidEuphoria.com</a>
|
Not Categorized, Please Help
|
|
