[Date Index] [Thread Index] [Author Index]

Testing for a non-constant variance

• To: mathgroup at smc.vnet.net
• Subject: [mg66100] Testing for a non-constant variance
• From: "john.hawkin at gmail.com" <john.hawkin at gmail.com>
• Date: Sat, 29 Apr 2006 03:40:59 -0400 (EDT)
• Sender: owner-wri-mathgroup at wolfram.com

Hello all,

I made a post a while back asking about testing a set of data for
stationarity.  While there are statistical tests that can be used to
detect a trend in a set of data, I am not aware of any that can detect
a non-constant variance.

One way of doing this would be to calculate the local standard
deviation for a series of points throughout the time series and compare
these to either the standard deviation of the whole data set, or the
average of these local standard deviations.  I am not sure exactly how
to implement this in an optimal way, ie what kind of cut-offs to use.

Does anyone know of any standard statistical tests that determine if a
data set has a non-constant variance, possibly ones that are
implemented in mathematica?  Or does anyone have any ideas on what a
good way to quantify my above method would be, in the general case?
Any thoughts would be greatly appreciated.