survivalgpu.simulate_dataset_batch

survivalgpu.simulate_dataset_batch(max_time, n_patients, list_covariates, batchsize=None, compress=True, seed=None)

Simulates a dataset of n_patients patients by concatenating smaller chunks.

Simulating all n_patients at once with simulate_dataset can be too slow or memory-intensive for large datasets. Instead, this function simulates successive chunks of batchsize patients (with a final, smaller chunk for the remainder, if any) and concatenates the results, offsetting the “patients” column of each chunk so that patient ids remain unique across the full dataset.

Parameters:
  • max_time (int) – the maximum follow-up time, passed to simulate_dataset.

  • n_patients (int) – the total number of patients in the final dataset.

  • list_covariates (list) – the covariates to simulate, passed to simulate_dataset.

  • batchsize (int, optional) – the number of patients to simulate per chunk. Defaults to n_patients, i.e. a single chunk.

  • compress (bool, optional) – whether to compress each chunk, passed to simulate_dataset. Defaults to True.

  • seed (int, optional) – random seed for the first chunk. Subsequent chunks use seed + i for reproducibility. Defaults to None.

Returns:

pandas.DataFrame –

the concatenated dataset, with n_patients distinct,

contiguous patient ids.