(From Sam(uel) Saunders [941028:2030 EDT])
I have been trying some variations on the changeover in Bill Powers' varint
simulation. I tried making the fixed delay inversely proportional to the
schedule interval, but did not come up with any set of parameters that looked
promising. A variant I tried that was more sucessful was to increase the fixed
delay after reinforcement. I got the sort of mild undermatching I would
hope to see. I need to experiment more to see how satisfactory this variant
may be.
In Conc VI VI, the propability of reinforcement remains constant for the
schedule in which the bird is responding, but is increasing in the other
schedule. At some point, no matter how lean the other schedule, the
probability
of reinforcement will be higher, and the bird should change over. When the
bird is responding in the leaner schedule, probability in the alternative will
quickly become higher, and this may be behind the need for a COD. I think this
may need to be included in the simulation, but I am not sure yet how to do it.
Sam
Samuel Spence Saunders, Ph.D.
sssaunders@delphi.com
···
--
Samuel Spence Saunders, Ph.D.
aa754@freent.buffalo.edu