ValueError: Domain error in arguments scipy rv_continuous

问题

I was trying to sample random variables subject to a given probability density function (pdf) with scipy.stats.rv_continuous:

class Distribution(stats.rv_continuous):
    def _pdf(self,x, _a, _c):
        return first_hitting_time(x, _a, _c)

where the function first_hitting_time is

#pdf of first hitting time of W_t + c*t on a. 
def first_hitting_time(_t, _a, _c=0.0):
    return _a/_t*np.exp(-0.5/_t*(_a-_c*_t)**2)/np.sqrt(2.0*np.pi*_t)

then I continue with

myrv= Distribution(name='hittingtime', a=0.002,b=30.0)
data3= myrv.rvs(size=10000, _a=1.0, _c=0.0)

and interpreter starts to complain-

Traceback (most recent call last):

  File "<ipython-input-246-71f67047462b>", line 1, in <module>
    data3= myrv.rvs(size=10000, _a=1.0, _c=0.0)

  File "C:\Users\ME\AppData\Local\Continuum\Anaconda2\lib\site-packages\scipy\stats\_distn_infrastructure.py", line 856, in rvs
    raise ValueError("Domain error in arguments.")

ValueError: Domain error in arguments.

it seems if I set _c to be some number larger than 0.0, it works fine, but not for _c less than 0.

I am a bit confused about this. Any help would be appreciated.

回答1:

From the documentation:

Subclassing

New random variables can be defined by subclassing the rv_continuous class and re-defining at least the _pdf or the _cdf method (normalized to location 0 and scale 1).

If positive argument checking is not correct for your RV then you will also need to re-define the _argcheck method.

It's not clear to me from your function what _a and _c represent, but it looks like you want to allow them to be negative.

See the default implementation in the source of _distn_infrastructure

来源：https://stackoverflow.com/questions/39420430/valueerror-domain-error-in-arguments-scipy-rv-continuous

标签

python

numpy

scipy