Revision history [back]
 | 1 | initial version |
You could fake it like this:
find_fit([(2,2),(3,2.9),(4,4.1)],(abs(a)+1)/abs(b)*x,variables=[x],initial_guess=(-2,-2))
So the best fit values of a,b are negative, but their absolute values are a "good fit" in your model.
