Thanks for the info.   We look forward to your results.  

By the way... our AgLeader FODD that some companies use probably "smooths" some of this site data more than other applications.  Instead of using a single "flow is delayed by 10 seconds" algorithm, we say flow is approximately delayed by 10 seconds.  But for any second, we use 10% of the yield value from 8 seconds, 20% from 9 seconds, 50% from 10 seconds, 20% from 11 seconds, and 10% from 12 seconds.   We don't believe that if you threw a colored ear of corn into the machine, that all of the color kernels would pass the flow sensor at 10 seconds... rather, the observations would taper up, and then taper off.   Whether you like this is a function of whether you believe in the plug flow model for combine dynamics, or you think the combine is a blender that is more appropriately modeled by a normal distribution for flow delay.   That is why I would have wanted your raw data, rather than processed shapefiles.