Skip to content

stats

def spearmanr(X) #

Calculate the spearman rank correlation between observer rankings.

Source code in src/mfpbench/stats.py 
11
12
13
14
15
16
17
18
19
20
21
22
def spearmanr(X: np.ndarray) -> np.ndarray:
    """Calculate the spearman rank correlation between observer rankings."""
    corr, _ = scipy.stats.spearmanr(X, axis=1)
    result: np.ndarray
    if isinstance(corr, float):
        result = np.eye(N=2, dtype=float)
        result[0, 1] = corr
        result[1, 0] = corr
    else:
        result = corr

    return result

def kendalltau(X) #

Calculate the kendall tau rank correlation between observer rankings.

Source code in src/mfpbench/stats.py 
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
def kendalltau(X: np.ndarray) -> np.ndarray:
    """Calculate the kendall tau rank correlation between observer rankings."""
    results: np.ndarray
    results = np.eye(N=len(X), dtype=float)

    idxs = range(len(X))
    for i, j in combinations(idxs, 2):
        x = X[i, :]
        y = X[j, :]

        corr, _ = scipy.stats.kendalltau(x, y)

        results[i, j] = corr
        results[j, i] = corr

    return results

def cosine(X) #

Calculate the cosine rank correlation between observer rankings.

Source code in src/mfpbench/stats.py 
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
def cosine(X: np.ndarray) -> np.ndarray:
    """Calculate the cosine rank correlation between observer rankings."""
    X_norm = np.zeros(shape=X.shape, like=X)
    for i, row in enumerate(X):
        mi = row.min()
        ma = row.max()
        X_norm[i] = 2 * ((row - mi) / (ma - mi)) - 1

    results: np.ndarray
    results = np.eye(N=len(X), dtype=float)
    idxs = range(len(X))
    for i, j in combinations(idxs, 2):
        x = X_norm[i, :10]
        y = X_norm[j, :10]

        corr = np.dot(x, y) / (norm(x) * norm(y))

        results[i, j] = corr
        results[j, i] = corr

    return results

def rank_correlation(x, *, method='cosine') #

Calculate rank correlation between observer rankings.

Will return the correlation between the two rankings, otherwise it will return a correlation matrix where each row represents a random variable.

PARAMETER DESCRIPTION
x

The rankings to calculate correlations for where each row

TYPE: ndarray

method

The method to use

TYPE: Literal['spearman', 'kendalltau', 'cosine'] DEFAULT: 'cosine'

RETURNS DESCRIPTION
ndarray

The correlation matrix
Source code in src/mfpbench/stats.py 
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
def rank_correlation(
    x: np.ndarray,
    *,
    method: Literal["spearman", "kendalltau", "cosine"] = "cosine",
) -> np.ndarray:
    """Calculate rank correlation between observer rankings.

    Will return the correlation between the two rankings,
    otherwise it will return a correlation matrix where each row represents a random
    variable.

    Args:
        x: The rankings to calculate correlations for where each row
        method: The method to use

    Returns:
        The correlation matrix
    """
    if not isinstance(x, np.ndarray):
        x = np.asarray(x)

    assert x.ndim >= 2

    measures = {"spearman": spearmanr, "kendalltau": kendalltau, "cosine": cosine}
    f = measures.get(method)
    if f is None:
        raise NotImplementedError(method)

    return f(x)