Code snippets

This is a collection of random utility functions that are out of scope for the package itself. These code snippets are placed in the public domain; you may use and modify them without attribution.

Progress reporting

These functions create a callback that prints a line for every 10% of estimation tasks done. They calculate the number of tasks by the x and lag array shape. The optional factor parameter can be used if you average several estimates together; the same callback can be passed to each estimation call.

def get_estimate_mi_callback(x, lags, factor=1):
    """Return a callback that prints a line for every 10% of estimation."""
    total_tasks = x.shape[1] * len(lags) * factor
    tasks_done = 0

    def callback(var_index, lag):
        nonlocal tasks_done
        tasks_done += 1

        cur_percent = (tasks_done / total_tasks) * 100
        prev_percent = ((tasks_done - 1) / total_tasks) * 100

        if cur_percent // 10 > prev_percent // 10:
            print(f"{datetime.now():%X}: {int(cur_percent):>3}% ({tasks_done}/{total_tasks})")

    return callback

def get_pairwise_mi_callback(x, factor=1):
    """Return a callback that prints a line for every 10% of estimation."""
    total_tasks = (x.shape[1] * (x.shape[1] - 1) // 2) * factor
    tasks_done = 0

    def callback(var_index, lag):
        nonlocal tasks_done
        tasks_done += 1

        cur_percent = (tasks_done / total_tasks) * 100
        prev_percent = ((tasks_done - 1) / total_tasks) * 100

        if cur_percent // 10 > prev_percent // 10:
            print(f"{datetime.now():%X}: {int(cur_percent):>3}% ({tasks_done}/{total_tasks})")

    return callback

To use the callback, pass the returned value to estimate_mi/pairwise_mi (or their ..._corr counterparts) as the optional callback keyword parameter:

y, covariates = ...
lags = [0, 1, 2, 3]
callback = get_estimate_mi_callback(covariates, lags)

mi = estimate_mi(y, covariates, lags, callback=callback)