Is the two-tailed pvalue correct in permutationTest?

[jodorning]

When I compare the output to the same test in Python permutation_test library, it seems that your PermutationTest outputs the most frequent mean rather than the two-tailed pvalue. The one-tailed p-values are similar to those from Python permutation_test.

SQL (BIGQUERY)

CREATE TEMPORARY FUNCTION permutation_test(sampleX ARRAY<FLOAT64>, sampleY ARRAY<FLOAT64>)
 RETURNS FLOAT64
 LANGUAGE js AS
"""
  return permutationTest(sampleX, sampleY, "two_side", 5000); //simple-statistics.alternative.permutationTest(sampleX: Array<number>, sampleY: Array<number>, alternative: string, k: number): number of permutations)
"""
OPTIONS (
 library="xxxx"
);

 SELECT permutation_test([2., 5, 3, 6, 7, 2, 5], [20., 5, 13, 12, 7, 2, 2])  

Two-sided P = 0.133

PYTHON

import permutation_test as p

var_control = [2, 5, 3, 6, 7, 2, 5]
var_treatment = [20, 5, 13, 12, 7, 2, 2]

p_value = p.permutationtest(
    data = var_control, 
    ref_data=var_treatment,
    n_combinations_max=5000)

Two-sided P = 0.0999