|
PureBytes Links
Trading Reference Links
|
The method most often used is to plot the change in the log of the price or the percent change in price. This is usually what is considered to be the normal distribution.
Change = 100 * (Close - Close[1]) / Close;
As I mentioned before, it tends to be a bit narrower in the middle of the distribution with fatter tails. So the probability of two-sigma or three-sigma events is actually higher than that predicted by the normal distribution.
Attached is a plot of the daily percent changes in GE for the past 750 trading days. It is roughly a normal distribution.
Bob Fulks
At 10:49 PM -0400 6/6/00, Ronald McEwan wrote:
>One of the members mentioned something that made an impression on me.
>"All Returns are Normally Distributed, until they are not....". I read
>the books and articles with the great plots of Log Normal and Normal
>Distribution Plots. So I tried to do this myself, and my plots were not
>even close to the ones in the articles and books. Must be something wrong
>with the way I did it? Well now I am not so sure. In the attached gif,
>you can see that the plot in the upper left looks like the Log Normal and
>Normal Distribution plots you see in the text books. The other plot is
>for 360 days of GE closing prices, using the exact same formulas for both
>plots. This does not look like the plots in the text books. My conclusion
>is that there is kind of "phase" shift of price distribution over time
>and changing market conditions . They can be very Normal, almost Normal,
>somewhat Normal and partly Log normal and most of the time very Log
>normal. Or something along those lines. By plotting these for myself I
>can see this transition unfold and will be better prepared for a
>potential "...until they are not..." event. I am still open to the fact
>that I have done this wrong and would welcome suggestions (with real
>examples) of how to do this.
Attachment Converted: "f:\eudora\attach\GE.gif"
|