arch.bootstrap.IndependentSamplesBootstrap

class arch.bootstrap.IndependentSamplesBootstrap(*args, **kwargs)[source]

Bootstrap where each input is independently resampled

Parameters
  • args (Union[ndarray, DataFrame, Series]) -- Positional arguments to bootstrap

  • kwargs (Union[RandomState, ndarray, DataFrame, Series]) -- Keyword arguments to bootstrap

data

Two-element tuple with the pos_data in the first position and kw_data in the second (pos_data, kw_data)

Type

tuple

pos_data

Tuple containing the positional arguments (in the order entered)

Type

tuple

kw_data

Dictionary containing the keyword arguments

Type

dict

Notes

This bootstrap independently resamples each input and so is only appropriate when the inputs are independent. This structure allows bootstrapping statistics that depend on samples with unequal length, as is common in some experiments. If data have cross-sectional dependence, so that observation i is related across all inputs, this bootstrap is inappropriate.

Supports numpy arrays and pandas Series and DataFrames. Data returned has the same type as the input date.

Data entered using keyword arguments is directly accessibly as an attribute.

To ensure a reproducible bootstrap, you must set the random_state attribute after the bootstrap has been created. See the example below. Note that random_state is a reserved keyword and any variable passed using this keyword must be an instance of RandomState.

Examples

Data can be accessed in a number of ways. Positional data is retained in the same order as it was entered when the bootstrap was initialized. Keyword data is available both as an attribute or using a dictionary syntax on kw_data.

>>> from arch.bootstrap import IndependentSamplesBootstrap
>>> from numpy.random import standard_normal
>>> y = standard_normal(500)
>>> x = standard_normal(200)
>>> z = standard_normal(2000)
>>> bs = IndependentSamplesBootstrap(x, y=y, z=z)
>>> for data in bs.bootstrap(100):
...     bs_x = data[0][0]
...     bs_y = data[1]['y']
...     bs_z = bs.z

Set the random_state if reproducibility is required

>>> from numpy.random import RandomState
>>> rs = RandomState(1234)
>>> bs = IndependentSamplesBootstrap(x, y=y, z=z, random_state=rs)

Methods

apply(func[, reps, extra_kwargs])

Applies a function to bootstrap replicated data

bootstrap(reps)

Iterator for use when bootstrapping

clone(*args, **kwargs)

Clones the bootstrap using different data.

conf_int(func[, reps, method, size, tail, ...])

type func

Callable[..., Union[ndarray, DataFrame, Series]]

cov(func[, reps, recenter, extra_kwargs])

Compute parameter covariance using bootstrap

get_state()

Gets the state of the bootstrap's random number generator

reset([use_seed])

Resets the bootstrap to either its initial state or the last seed.

seed(value)

Seeds the bootstrap's random number generator

set_state(state)

Sets the state of the bootstrap's random number generator

update_indices()

Update indices for the next iteration of the bootstrap.

var(func[, reps, recenter, extra_kwargs])

Compute parameter variance using bootstrap

Properties

index

Returns the current index of the bootstrap

random_state

Set or get the instance random state