Hacker's Diet has a good section explaining rolling averages. It's
really good for indicating trends instead of daily fluctuations.
Here's the start of why to use them:
http://www.fourmilab.ch/hackdiet/www/subsection1_2_4_0_4.html#SECTION0240400000000000000
What I really like is the Floats and Sinkers charts a few pages later.
And here's the explanation of the math:
http://www.fourmilab.ch/hackdiet/www/subsection1_4_1_0_8.html#SECTION0410800000000000000
And that's about all I've got. I know next to nothing about statistics.
-Kevin
On Mon, 12 Jul 2004 12:33:42 -0400, Jerry Johnson
<[EMAIL PROTECTED]> wrote:
> To smooth a curve, you can take a rolling average (by taking the average of the data point and a couple of datapoints adjacent)
>
> for example
>
> 3day - take the day before, the day, and the day after, and average them (db+d+da)/3
>
> 1day 3day 5day
>
> s 2
> m 2 1.66
> t 1 1.66 1.8
> w 2 1.66 2.0
> t 2 2.33 2.2
> f 3 2.66 2.8
> s 3 3.33 3.0
> s 4 3.33 3.2
> m 3 3.33 5.4
> t 3 6.66 5.4
> w 14 10 5.0
> t 3 6.33 5.0
> f 2 2.66 4.8
> s 3 2.33
> s 2
>
> You'll see that the 5 day is mush smoother a curve than the 3day, which is smoother than the spiky day version.
>
> Does this help?
> Jerry Johnson
>
> >>> [EMAIL PROTECTED] 07/12/04 12:06PM >>>
> What do you mean by moving average? Just show the last N days in a
> range you mean? (That actually would help me.)
>
>
>
[Todays Threads] [This Message] [Subscription] [Fast Unsubscribe] [User Settings] [Donations and Support]
