Quick question on coefficient of variation

Floyt · 4th December 2007, 08:26 PM

I have a quick question about the equivalence of CV and variance - need to scram in 45mins and would dearly love to get this sorted beforehand, and can't easily find an answer right now - :

Seeing that the CV is just standard deviation divided by sample mean, can I subject it to the same kind of transformations that I could apply to the standard deviation or the variance? I need to compute a model-averaged CV over a set of models, and the technique I have available deals with variances, whereas I only have access to the CV output of my models (and no easy way to get at the mean and thus calculate the variance, either).
Seems to me that since the CV differs from the variance only by application of a constant, the same formula that gives me an averaged variance should give me a valid averaged CV when I plug it in - right?

Help very much appreciated

Jeff Corey · 4th December 2007, 09:09 PM

I don't think so because the Sd squared is the variance. So the constant would not appear to work. You need the means to get the difference. Sorry, no help.

Floyt · 4th December 2007, 10:35 PM

Darn

Hmm, maybe transform the formula for sd, and use THAT with the CV?

4th December 2007, 08:26 PM	#1
Floyt Chordate Join Date: Apr 2003 Location: Cape Town! Not mugged yet. Looking for chameleons. Posts: 1,428	Quick question on coefficient of variation I have a quick question about the equivalence of CV and variance - need to scram in 45mins and would dearly love to get this sorted beforehand, and can't easily find an answer right now - : Seeing that the CV is just standard deviation divided by sample mean, can I subject it to the same kind of transformations that I could apply to the standard deviation or the variance? I need to compute a model-averaged CV over a set of models, and the technique I have available deals with variances, whereas I only have access to the CV output of my models (and no easy way to get at the mean and thus calculate the variance, either). Seems to me that since the CV differs from the variance only by application of a constant, the same formula that gives me an averaged variance should give me a valid averaged CV when I plug it in - right? Help very much appreciated
	__________________ They had no god; they had no gods; they had no faith. What they appear to have had is a working metaphor. - Ursula K. Le Guin, "Always Coming Home" Last edited by Floyt; 4th December 2007 at 08:27 PM.

4th December 2007, 10:35 PM	#3
Floyt Chordate Join Date: Apr 2003 Location: Cape Town! Not mugged yet. Looking for chameleons. Posts: 1,428	Darn Hmm, maybe transform the formula for sd, and use THAT with the CV?
	__________________ They had no god; they had no gods; they had no faith. What they appear to have had is a working metaphor. - Ursula K. Le Guin, "Always Coming Home"

4th December 2007, 09:09 PM	#2
Jeff Corey New York Skeptic Join Date: Aug 2001 Posts: 13,795	I don't think so because the Sd squared is the variance. So the constant would not appear to work. You need the means to get the difference. Sorry, no help.