This dialog provides rudimentary help for planning a pilot
study for estimating an error SD (sigma).  If the pilot
study is too small, the estimate could be way too small,
causing us to seriously underestimate the required sample
size (N) for the main study.

Thus, the criterion employed in this dialog is the risk of
underestimating N by a specified percentage 100*P.  This
is approximately equal to the risk of underestimating
sigma (pretty much regardless of what criterion -- power,
CI width, etc. -- is used for sample-size determination).
In a normal setting, this risk is a simple calculation on
the chi-square distribution: risk = Pr{Q < (1-P)nu}, where
Q is chi-square with nu d.f.

The dialog has three sliders -- for 100P, for risk, and
for nu.  If you vary P or nu, the risk is recomuted based
on the above formula.  If you vary risk, nu is recomputed.
(If the solve method fails, simply set the d.f. so the
risk is in the ballpark, and try again.)
