"newshound" wrote in message
...
I'd say the best approach would be to find a transformation (eg y=log(x)
or y=sqrt(x)) which gives a good, well-correlated straight line. Then
extrapolate that and do an inverse transformation (eg antilog or
x-squared) on the predicted value. Obviously the more data points you
have, the better prediction you can make and the better you can construct
a least-squares regression line for extrapolation and then
back-transformation.

Sorry, but trying to get the "right" straight line from three points is
pretty meaningless. Batteries are intrinsically non-linear. You can (sort
of) rely on an exponential for radioactive decay when you are starting
with maybe 10^20 atoms. Batteries are more complicated because you have
competing processes.

Yes, I never suggested that the data would be linear. I'd be highly
surprised if it was anything like linear. But if you can find a way of
transforming the non-linear data so it becomes linear, then you can work out
the best-fit line and predict future (transformed) values, from which you
can back-transform to get the corresponding real value. Easier to do it that
way that to try to fit a flexicurve to a curved line so you can predict an
out-of-range value (extrapolating) as opposed to predicting an in-range
value (interpolating).