ehrdata.dt.ehrdata_blobs

ehrdata.dt.ehrdata_blobs(*, n_variables=11, n_cat_vars=0, n_categories=None, n_centers=5, cluster_std=1.0, n_observations=1000, base_timepoints=100, random_state=0, sparse=False, sparsity=0.9, variable_length=False, time_shifts=False, seasonality=False, irregular_sampling=False, missing_values=0.0, layer='tem_data')

Generates time series example dataset suited for alignment tasks.

Parameters:

layer str (default: 'tem_data')

The name of the layer to store the data in. If not specified, uses X.

n_variables int (default: 11)

Dimension of feature space.

n_cat_vars int (default: 0)

Number of categorical variables.

n_categories list[int] | None (default: None)

List of cardinalities for each categorical variable.

n_centers int (default: 5)

Number of cluster centers.

cluster_std float (default: 1.0)

Standard deviation of clusters.

n_observations int (default: 1000)

Number of observations.

base_timepoints int (default: 100)

Base number of time points (actual may vary per observation).

random_state int | Generator (default: 0)

Determines random number generation for dataset creation.

sparse bool (default: False)

Whether to use sparse matrices.

sparsity float (default: 0.9)

Target sparsity level when sparse=True.

variable_length bool (default: False)

Whether observations have different time series lengths.

time_shifts bool (default: False)

Whether to add time shifts between similar observations.

seasonality bool (default: False)

Whether to add seasonal patterns to time series.

irregular_sampling bool (default: False)

Whether sampling intervals vary between observations.

missing_values float (default: 0.0)

Fraction of random missing values in time series.

layer

The name of the layer to store the time series data in.

Return type:

EHRData

Examples

>>> import ehrdata as ed
>>> edata = ed.dt.ehrdata_blobs(
...     variable_length=True, time_shifts=True, seasonality=True, irregular_sampling=True
... )

results in a dataset like:

Categorical variables can be generated with different cardinalities per variable (e.g. 2, 3, 4 categories). Different clusters (groups) can also exhibit different category distributions: