CUP is dead, may it forever R. I. P.
Old stuff and memories are not to be discarded, just filed away so that future generations may see the road we traveled.
This is my first use of this site and I cannot find a means to attach a megabyte graphic file. I have a regression relating PSI to CUP based upon an ANSI/SAAMI 2015 data set. This data set is an extension of the ANSI 2002 data set that denton regressed in previous years. My regression is also not based directly upon scientific measurements, but upon maximum ratings that are generously guard-banded above a set of scientific measurements. For this reason, do not have large expectations of it's precision in PSI to CUP conversions. In general, conversions from my regressions are not greatly different than the linear conversions from denton's previous work.
My regression is not linear, despite its looks. I put the PSI data on the X axis because those measurements are more precise, however, with the guard-banding and round off, the PSI data is probably not significantly better than the CUP data. The regression is curved and very nearly parabolic, however, my chosen log function may be written in closed form as Y = f(X), and X = f(Y) which cannot be done with a parabolic (2nd order polynomial) function. It is non-linear because the X-scale is logarithmic with an offset, thereby including the third degree of freedom. A straight-line regression was required in order to establish the confidence intervals shown. In the confidence interval expression, due to the subtraction (difference of similar large numbers), higher than normal precision is required in this calculation. I have included all of the useful statistical information and equations on the graphic file.
Last edited by Prunty; 04/11/19.