def estimate_slope()

in spotify_confidence/analysis/frequentist/confidence_computers/z_test_linreg_computer.py [0:0]

• 26 lines of code
• 6 McCabe index (conditional complexity)

def estimate_slope(df, **kwargs: Dict) -> DataFrame:
    if kwargs[FEATURE] not in df:
        return df

    def col_sum(x):
        return reduce(lambda x, y: x + y, x)

    def dimension(x):
        return x.shape[0] if isinstance(x, np.ndarray) and x.size > 1 else 1

    k = df[kwargs[FEATURE_SUMSQ]].apply(dimension).iloc[0]

    XX0 = np.zeros((k + 1, k + 1))
    XX0[1 : (k + 1), 1 : (k + 1)] = col_sum(df[kwargs[FEATURE_SUMSQ]])
    XX0[0, 0] = col_sum(df[kwargs[DENOMINATOR]])
    XX0[0, 1 : (k + 1)] = col_sum(df[kwargs[FEATURE]])
    XX0[1 : (k + 1), 0] = col_sum(df[kwargs[FEATURE]])

    Xy0 = np.zeros((k + 1, 1))
    Xy0[0,] = col_sum(df[kwargs[NUMERATOR]])
    Xy0[1 : (k + 1),] = np.atleast_2d(col_sum(df[kwargs[FEATURE_CROSS]])).reshape(-1, 1)

    try:
        b = np.matmul(np.linalg.inv(XX0), Xy0)
    except np.linalg.LinAlgError:
        b = np.zeros((k + 1, 1))
    out = b[1 : (k + 1)]
    if out.size == 1:
        out = out.item()

    outseries = Series(index=df.index, dtype=df[kwargs[FEATURE]].dtype)
    df[REGRESSION_PARAM] = outseries.apply(lambda x: out)
    return df