Adding bootstrap sampling to vp2

Created by: wilsonmr

Hi,

I've been messing around trying to add a bootstrap sampling method to vp2 but have gotten a bit stuck.

Basically I would like to get something similar to the output of results however instead of the theory results coming from the original PDF, being a monte carlo sample of those replicas.

Just to test the idea I added something like this:

def bootstrap_test(results):
    """try to bootstrap values"""
    data_result, th_result = results
    alldata, central, npoints = abs_chi2_data(results)
    th_sample  = th_result
    th_sample_i = np.floor(np.random.rand(len(th_result._rawdata[0]))*len(th_result._rawdata[0])).astype(int)
    th_sample._rawdata = th_result._rawdata[:, th_sample_i]
    results = data_result, th_sample
    alldata_sample, central_sample, npoints = abs_chi2_data(results)
    return alldata.data.mean(), alldata_sample.data.mean()

which does give what I want for alldata, however since I'm just editing the _rawdata property of th_result here, things like central_value property remains untouched so this appears to be the wrong way to go about it.

I think the idea we were thinking of would be to have a bootstrapping function which can take any function in say results.py and calculate it multiple times for different monte carlo samples of replicas.. so somehow feed a different results in each time which is the new results from sampling. Is something like this possible?