Heidi,

What Paul outlined is a good start, but if you want to run a sensitivity
analysis where you vary cultivar coefficients, you would still need a way
to provide keyboard input. An lternate approach is to create dummy
cultivars with different combinations of cultivar coefficients.
To optimize on this, I expect you could do something like this:

1. Run initial set of generic cultivars (say 200, which would only take a
few minutes)
2. Identify best candidate values
3. Create a new cultivar file with values iterated around those of the
best cultivars
4. Find best or set of best ... and go back to step three

My main concern is that unless you have a lot of very clean data, you may
find that mutliple cultivar types give you very similar results. Thus,
rather than finding a single unique soultion, you will find several
options that look equally promising. Automated coefficient generation is
always worth trying, but fitting that is guided by field knowledge of the
genotypes may ultimately provide  more reliable. In common bean, two key
issues to look at are the growth habit and the expected seed mass under
good production conditions.

Although I am not working with cultivar optimization right now, I often
use long DOS-type batch files (*.BAT) to control model runs.

Best regards,
Jeff

Jeffrey W. White
ALARC, USDA-ARS
21881 North Cardon Lane
Maricopa, AZ 85239 USA
Tel. 1-520-316-6368
Fax 1-520-316-6330
[log in to unmask]
NOTE NEW mailing address and phone as a result of our reorganization and
relocation in Feb. 2006.