FItting a dose-response curve, Prism's results seem wrong.

There is a bug in Prism version 4.00-4.02 and 4.0a-4.0b. It will appear only if you fit the standard (not variable slope) dose response curve and share one or more parameters between data sets. To bypass the bug, use the variable slope dose-response curve instead, and fix the Hill Slope to -1. This was fixed in 4.03/4.0c.