Just to satisfy my own curiosoty Dave: Why the fudge factors in your old cubic and Novotny linear calculations? Is it just that those modified values are just closer than the New cubic and New Linear calcs so you call that good?
Yes. Trying to fit the line better, assuming the curvature is correct but vertical placement is not. Call my versions "the Taylor formulae" if you prefer. Originally, I had applied fudge factors to Terrill's formulae based on Terrill's own analysis... but found those fudgers unnecessary to fit my own data set. The reasons for doing so were sound enough for me to go ahead and use fudgers for other formulae where necessary. See writeup from Terrill here, where he suggests his own New Cubic averages high by about 0.0007, but the Old Cubic was low by about 0.0017... so I added those fudge factors into my own spreadsheet originally right off the bat, then tweaked from there based on the line in my spreadsheet marked "New ABS Method" with a goal of zero of course, which unlike Terrill's "Mean discrepancy" which blends negative and positive values, my method uses absolute value distances from the goal line, yadda yadda. Anyway...... here's Terrill's own analysis of the accuracy of his own formulae:
http://seanterrill.com/2011/04/07/refractometer-fg-results/I guess you could say, I am attempting to normalize the playing field, instead of allowing each formula to average too high or too low. Let's assume the curvature is correct but for some reason the people taking measurements aren't calibrating their hydrometer accurately, or whatever, which would result in a consistently high or low correlation.
There you have it. I'm finally done editing this post (edited it about 6 times in the past 8 minutes).