Feature to retrieve all sampled results so that a histogram can be obtained

[Boxylmer]

Please forgive me if this feature exists, but is there a way to easily pull a list of the resampled outputs so that I can create a histogram of the values predicted? I'd like to get a convincing visual figure that shows I've used enough samples.

[juliangehring]

Are you looking for a way to collect 1) the resampled input data or 2) the outputs of the statistic function that was applied to the resampled data?

[Boxylmer]

The latter. If I hand Bootstrap an input of something like [1.2, 3.2, 1.1, 1.0, 0.9], and I get back a bunch of resampled values from the model being bootstrapped (say n=4) as [2.1, 2.2, 2.1, 2.0], I want that vector of outputs so that I can verify a histogram of the results "looks" correct. Is this already possible with the current exported functions?

[juliangehring]

Yes, you can get the output statistic of each resampling with the straps method (it should probably have a more meaningful name):

bs = bootstrap(median, [1.2, 3.2, 1.1, 1.0, 0.9], BasicSampling(5));
straps(bs)

which yields for example:

([1.0, 1.2, 3.2, 1.1, 1.2],)

The output is a tuple where each tuple element corresponds to the respective output values of the statistic function. In this case, since median only returns a single value per resampling, the first tuple element contains all the medians estimated from each resampling.

[Boxylmer]

Thanks! I missed it completely. Now I can just plot a histogram of straps(bs) and hopefully I'll see something bell shaped :)

[skleinbo]

Can straps be regarded as API? If yes, it should be documented.

[Boxylmer]

(I'm all for this as well!)