Priors

Specifying prior distributions

kamphir uses the statistical distribution objects in SciPy.stats to specify prior distributions. The continuous distributions are all derived from a base object that takes two standard arguments, loc and scale. For example, a uniform distribution over (100, 200) can be specified as uniform(loc=100, scale=100). Specifically, this is the string that would be passed to the prior keyword in a kamphir settings JSON file. All distribution specifications become prefixed by scipy.stats in kamphir.

A good way to check a prior distribution is to compute the mean, standard deviation, and draw a number of random deviates in Python:

>>> from scipy.stats import uniform
>>> f = uniform(loc=100, scale=100)
>>> f.mean()
150.0
>>> f.std()
28.867513459481287
>>> [f.rvs() for i in range(10)]
[115.5179616986701, 191.5066287995557, 199.25118264899623, 187.45248482228178, 184.82615419309076, 154.14140984337394, 195.20574006635417, 190.9157583695232, 115.86909627658044, 188.8181812366821]

The random variates can then be passed into R to generate a histogram.