convert_responses_to_kernel_sign

def convert_responses_to_kernel_sign(responses):
    """Converts dichotomous responses to the appropriate kernel sign.

    Takes in an array of responses coded as either [True/False] or [0/1]
    and converts it into [+1 / -1] to be used during parameter estimation.

    Values that are not 0 or 1 are converted into a zero which means these
    values do not contribute to parameter estimates.  This can be used to 
    account for missing values.

    Args:
        responses: [n_items x n_participants] array of response values

    Returns:
        the_sign: (2d array) sign values associated with input responses
    """
    # The default value is now 0
    the_sign = np.zeros_like(responses, dtype='float')

    # 1 -> -1
    mask = responses == 1
    the_sign[mask] = -1

    # 0 -> 1
    mask = responses == 0
    the_sign[mask] = 1

    return the_sign