Raghunathan,

A polynomial f(x) of order n will exhibit n-1 inversions wherein df/dx = 0. In other words, it will exhibit n-1 "humps" or "valleys" wherein the curve changes direction. His data looks pretty linear, but his error is clearly a 2nd order polynomial (in the domain of x = [0,...,600]).

Nagarajan,

Try doing a 2nd order polynomial fit of ERR CT (Y) as a function of ACTUAL CT (X). I just did one manually (graphing the ERR CT and a fitted function F(X) together). By playing with the A0, A1, and A2 variables, for the function F(X) = A0 + A1*X + A2*X^2 (X = ACTUAL CT), I got it to A0 = 19, A1 = 9.6E-4, and A2 = 8.2E-10 fairly quickly. Plug it into your spreadsheet, plot the two functions together and twiddle the A values until you get it fitted as nicely as you want.