Replying to LO27848 --
In a message dated 2/15/02 7:58:12 PM Eastern Standard Time,
amdelange@gold.up.ac.za writes:
> Scott -- 12 400 000
> Scott fun -- 1 070 000
> Scott fun statistics -- 73 200
> Scott fun statistics people -- 46 900
> Scott fun statistics people organisations --1 630
> Scott fun statistics people organisations "statistical detail" 2.
>
> You are lucky to get so many hits -- 2 to be precise. It shows what one
> can do with stats! ... grin ...
LO Group:
Actually, what has been generated here is not a statistic, but a
frequency count. If several different engines were to be used, we could
create a frequency distribution.
A statistic is merely an "estimate" of a parameter. Of course, all
measurements contain error, as any score ("x") is generated from two
elements, that of (1) the "true" score, (always unknown) and (2) +/-
error, (x = t + e). If we measure all the defined elements in a defined
population, the arithmetic reduction is called a parameter. Because this
is usually not possible, or perhaps desirable, we normally take a random
sample of elements within the population, and thus encounter "sampling
error," another important source of error in descriptive statistics.
Because we assume that "error" randomly distributes itself about the
"true" score with repeated samples, multiple measurements are desirable.
When we arithmetically reduce scores from the sample, we produce a
statistic. Thus, a statistic is an "estimate" of a parameter. When
properly understood and calcuclated, a statistic is very useful.
However, it is only useful if one understands the standard error of
the measurement, the sampling distribution error, and sample size relative
to the population. In addition, all of this simply provides estimates of
differing probabilities at different confidence levels.
Parametric and non-parametric statistics: such wonderful tools, I
can never do just one!
Warm regards, David
David L. Hanson, Ph.D.
Consulting Psychologist
Charlotte, NC
PS: Statistics don't "lie." Statistics are most often misunderstood and,
therefore, abused.
--Learning-org -- Hosted by Rick Karash <Richard@Karash.com> Public Dialog on Learning Organizations -- <http://www.learning-org.com>
"Learning-org" and the format of our message identifiers (LO1234, etc.) are trademarks of Richard Karash.