I finally got around to plot the dose/response curve (Boltzmann) for the recommended corn population versus expected (historic) yield. See this thread:
http://talk.newagtalk.com/forums/thread-view.asp?tid=455088&mid=3697440#M3697440
I have not yet seen how well SMS (which I am fairly certain is a modified ArcGIS core) likes these types of exponentials. Recall the Boltzmann for a dose response takes the form:
y = A2 + (A1-A2)/(1 + exp((x-x0)/dx)), where x is your historic yield A1 A2 x0 dx 19000 35500 123.6673 18.58726
Unforced A2 fit : A1 A2 x0 dx 19000 34957.54782 121.18642 16.71197 Info on Boltzmann "one site competition" or "sigmoidal dose-response": http://www.originlab.com/doc/Origin-Help/Boltzmann-FitFunc
I have attached a summary sheet as well as my Origins file for the forced fit (A1 and A2 force through parameters of min and max populations). The EC50 is the most critical I think; mid-point slope also important. Please comment and provide more source data. I will continue to post my parametric prescription generators. I am already not completely satisfied with ranges 150-200, 75-110 could also use some more data. There is interpolation inherent in this model - do not take it as gospel. It needs a primary data source (prospective trials).
"Multi-site cooperative binding" fits might be better, but more data is needed. Biologically, this might make more sense.
Attachments ----------------
Corn_pop_Boltzman.opj (229KB - 139 downloads)
Boltzman Fit Corn Population Prescription.doc (359KB - 147 downloads)
|