cbellei / abyes Goto Github PK

AttributeError: module 'abyes' has no attribute 'AbExp'

AttributeError: module 'abyes' has no attribute 'AbExp'

I ran the following code many times using your aByes library. Sometimes it reports A or B is winner. This is very odd because both has the mean of 0.4 and only runs for 100 data points. Shouldn't the result of "inconclusive" or "effectively equivalent" be expected?

data = [np.random.binomial(1, 0.4, size=100), np.random.binomial(1, 0.4, size=100)]
exp = ab.AbExp(method='analytic', rule='loss', decision_var='lift', toc=0.002, plot=True)
exp.experiment(data)

Hi @cbellei,

Thanks for an amazing load of work here. Really impressive.

While going over the code I saw this:

sigma_a_rvs = np.sqrt(a_rvs * (1 - a_rvs))
sigma_b_rvs = np.sqrt(b_rvs * (1 - b_rvs))

at: https://github.com/cbellei/abyes/blob/master/abyes/ab_exp.py#L148

Shouldn't the n_a in the formula as well?
As for binomial distribution
standard deviation (σx) is sqrt[ n * P * ( 1 - P ) ].

Thanks in advance. I'm happy to send a PR, but wanted to check first if I'm correct.