…weighting the sampled result of a distribution, we use the weighting to pick a distribution from which to sample.
Here is an example comparing the results from the current code (on top) and the new code (on bottom). As with the LSFF code, the mean remains similar but the tails get longer.
In [49]: pd.Series(ppf).describe()
Out[49]:
count 1000.000000
mean 28.778341
std 21.952865
min -0.729606
25% 10.471523
50% 24.585799
75% 43.283797
max 109.786157
dtype: float64
In [50]: pd.Series(ppfA).describe()
Out[50]:
count 1000.000000
mean 28.577141
std 22.632246
min -2.437593
25% 10.733608
50% 25.078260
75% 42.765163
max 228.270264
dtype: float64
Known issues: this isn't tied into the Common Random Number system in the framework. I like the use of np.random.choice() -- it seems fairly intuitive. What could be done to make it play nicely with the rest of the framework?
…weighting the sampled result of a distribution, we use the weighting to pick a distribution from which to sample.
Here is an example comparing the results from the current code (on top) and the new code (on bottom). As with the LSFF code, the mean remains similar but the tails get longer.
Known issues: this isn't tied into the Common Random Number system in the framework. I like the use of np.random.choice() -- it seems fairly intuitive. What could be done to make it play nicely with the rest of the framework?