I found John Denker's e-mail on graphical analysis illuminating and
the ideas he presented there stimulating and thought provoking.
However, I do have a problem with the development of the curve fit
for the resistance data. I don't doubt that the arithmetic done in
the analysis leads to the expressions John quotes, viz,
log(R) = 0.1136/sqrt(T) + 0.327 and/or log(R) = 0.1908 T^(-.85) +
0.1433.
What worries me is the large difference in precision between the
initial, mostly 1-digit, data and the 4,3, and 2 digit coefficients
that come out of the number-crunching.
Is the technique being used appropriate to such low precision initial
data?
If the answer is that the original data somehow or other had much
greater precision implied than the 1-digit values quoted, well, I'm
afraid I don't believe it. Is the value of the resistance or the
resistor at 3.000K really 3.000 ohms? I don't think so!
I know John Mallinckrodt wrote "Assuming that your data is good".
Again, what do the two Johns mean by "good" data. Is the amazing
coincidences of the zeroes I wrote above implied?