Stan Brown <[EMAIL PROTECTED]> wrote in
news:[EMAIL PROTECTED]: 

> The empirical rule tells you that _about_ 2.5% of the data will lie 
> above z = 2 in a normal distribution, not that _exactly_ 2.5% will 
> lie above. The smaller the population, the less closely it will 
> approach an ideal normal distribution and therefore the less good 
> that empirical rule will be. For 500 students the approximation 
> should not be too bad, _if_ the underlying distribution is 
> normal.(*) So your calculation correctly tells you that _about_ 12.5 
> students will have scores over 83. You would simply state this as 
> about 12 or 13 students.

Another way to put it is that the OP's original distribution is discrete, 
but he's approximating it with a continuous distribution, and the area 
under the appropriate part of the curve of a continuous distribution 
doesn't have to evenly divide that of a discrete one.

> (*) I'm a little uneasy about that "mound shape distribution" in the 
> problem statement. I hope you understand, Mel, that the empirical 
> rule does not apply to every mound-shaped distribution but only to 
> normal distributions. Other mound-shaped distributions will differ 
> from the empirical rule to a greater or lesser extent.

Same here.  "Mound shaped" to me merely implies that the distribution is 
unimodal, in which the appropriate "empirical rule" is really the Camp-
Meidel inequality; at most 1/(2.25*K^2) of the data will lie more than K 
standard deviations from the mean, regardless of the nature of the 
distribution (though even then it's an approximation when dealing with a 
discrete distribution).  In the OP's case, that would be 11%.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to