Closed rueuntal closed 8 years ago
About commit #3502071: using fsolve()
instead of bisect()
alleviates the need to define an upper bound for p, which is a bit tricky mathematically. Using @davharris 's test SAD the new version returns a slightly higher (~2e-6) likelihood compared to the old version. It would be nice to have more test cases though.
Looks good to me from a read through. Thanks! Agree it would be nice to add some more tests.
Copied from
mete.py
, with added logit transformation for parameter p.