xgbse.converters.convert_to_structured

Converts data in time (T) and event (E) format to a structured numpy array. Provides common interface to other libraries such as sksurv and sklearn.

Parameters:

Name	Type	Description	Default
T	np.array	Array of times	required
E	np.array	Array of events	required

Returns:

Type	Description
np.array	Structured array containing the boolean event indicator as first field, and time of event or time of censoring as second field

Source code in xgbse/converters.py

def convert_to_structured(T, E):
    """
    Converts data in time (T) and event (E) format to a structured numpy array.
    Provides common interface to other libraries such as sksurv and sklearn.

    Args:
        T (np.array): Array of times
        E (np.array): Array of events

    Returns:
        np.array: Structured array containing the boolean event indicator
            as first field, and time of event or time of censoring as second field
    """
    # dtypes for conversion
    default_dtypes = {"names": ("c1", "c2"), "formats": ("bool", "f8")}

    # concat of events and times
    concat = list(zip(E.values, T.values))

    # return structured array
    return np.array(concat, dtype=default_dtypes)

xgbse.converters.convert_y

Convert structured array y into an array of event indicators (E) and time of events (T).

Parameters:

Name	Type	Description	Default
y	structured array(numpy.bool_, numpy.number	Binary event indicator as first field, and time of event or time of censoring as second field.	required

Returns:

Type	Description
T ([np.array, pd.Series])	Time of events E ([np.array, pd.Series]): Binary event indicator

Source code in xgbse/converters.py

def convert_y(y):
    """
    Convert structured array y into an array of
    event indicators (E) and time of events (T).

    Args:
        y (structured array(numpy.bool_, numpy.number)): Binary event indicator as first field,
            and time of event or time of censoring as second field.

    Returns:
        T ([np.array, pd.Series]): Time of events
        E ([np.array, pd.Series]): Binary event indicator
    """
    event_field, time_field = y.dtype.names
    return y[event_field], y[time_field]