Centre county Pennsylvania, USA | Mark, thanks for compliment. CCYM historical forecast accuracy for national corn yields is indeed surprising. I think its a keeper (attached chart). Basically, this CCYM weights each of the 5 conditions reported weekly by the NASS then computes a yield contribution for each of those 5 conditions and sums the yield contributions from all conditions. Coefficients (twiddle factors) are used in that weighting. CCYM then predicts a final yield each week, and declares its prediction for week of Jul 20 to be the CCYM forecast of final corn yield for that year.
I don't know how USDA NASS makes their Aug 12 corn yield forecast, but CCYM corn yield forecast seems to be as accurate as NASS.... and is 3 weeks earlier than NASS.
I'm working on a CCYM for state corn yields and a CCYM for national soybean. When that work is producing surprising results, I'll post here on NAT.
(national _corn_yield_forecast_errors.png)
Attachments ---------------- national _corn_yield_forecast_errors.png (61KB - 33 downloads)
|