Date: Tue, 23 Feb 1999 18:11:29 0300
ReplyTo: "Hector E. Maletta" <hmaletta@OVERNET.COM.AR>
Sender: "SPSSX(r) Discussion" <SPSSXL@UGA.CC.UGA.EDU>
From: "Hector E. Maletta" <hmaletta@OVERNET.COM.AR>
Subject: Re: Coefficient of Variation
ContentType: text/plain; charset=usascii
Its the ratio of standard deviation to the mean. More than 'precision'
it reflects the relative dispersion of data around the mean. Besides, in
a normal distribution 95% of the cases lie within 1.96 standard
deviations around the mean, thus if the coefficient is, say, 0.15, and
the distribution is normal, then about 95% of the cases will lie in the
interval delimited by (10.30)*mean and (1+0.30)*mean. Many
psychological, medical and biological variables are known or expected to
have a nearly normal distribution of frequencies, and thus this concept
may be handy. Sociological and economic variables, though, have usually
nonnormal distributions.
All this, of course, is not to be confused with the distribution of
sampling errors, which is always expected to be normal if the samples
are randomly selected and of sufficient size. The standard deviation of
the sampling distribution (variability among many hypothetical samples
of the same size) is usually estimated as the standard deviation
observed in your sample, divided by the square root of the sample size.
But in this context the notion of a coefficient of variation, though
feasible, is not commonly used.
Hector Maletta
Universidad del Salvador
Buenos Aires, Argentina
shailendra@CANADA.COM wrote:
>
> Recently I came across a term called coefficient of variation, which measures the precision of the estimated data.
>
> My curiosity how important it is know the coefficient of variation in survey analysis, how it is calculated at the sample level, and is there any % benchmark for accepting data for analysis.
>
> Any reference or comments are thankfully welcome.
>
> 
> Get free personalized email from Canada.com at http://www.ehmail.com
