# How to quantify the volitility of a time series data?

I am currently working on a time series data and I would like to quantify how volatile it is.

Here volatile I mean how "shaky" the series is.

If the series is smooth than it is not volatile.

I have an idea to solve this problem, but it is kind of inconvenient.

The idea is to first do a regression/smoothing on the series. Then compute the sum of squared error between the smoothed series and the original series.

Any other better idea or and references suggest me to have a look?

-
How will you do the regression/smoothing. If you use a polynomial of arbitrarily high degree, you can get a smooth curve which is exactly as "volatile" as your time-series. –  utdiscant Jul 25 '12 at 5:35
The over fitting/training is also a problem. I am thinking to use spline but I still need to read more about it. –  Rein Jul 25 '12 at 6:23
BTW: volatility in finance is more or less standard deviation. –  user2468 Jul 25 '12 at 6:26