experiment.py 文件源码

def calc_pvalues_pairwise(self, label):
        """
        Calculate pvalues for each variant in each pair of Conditions.
        """
        if self.check_store("/main/{}/scores_pvalues".format(label)):
            return

        data = self.store['/main/{}/scores'.format(label)]

        cnd1_index = list()
        cnd2_index = list()
        values_index = list()

        values_list = ["z", "pvalue_raw"]

        condition_labels = data.columns.levels[0]
        for i, cnd1 in enumerate(condition_labels):
            for cnd2 in condition_labels[i + 1:]:
                cnd1_index.extend([cnd1] * len(values_list))
                cnd2_index.extend([cnd2] * len(values_list))
                values_index.extend(sorted(values_list))
        columns = pd.MultiIndex.from_tuples(zip(cnd1_index, cnd2_index,
                                                values_index),
                                            names=["condition1", "condition2",
                                                   "value"])

        idx = pd.IndexSlice
        result_df = pd.DataFrame(np.nan, index=data.index, columns=columns)
        for i, cnd1 in enumerate(condition_labels):
            for cnd2 in condition_labels[i + 1:]:
                result_df.loc[:, idx[cnd1, cnd2, 'z']] = \
                    np.absolute(data.loc[:, idx[cnd1, 'score']] -
                                data.loc[:, idx[cnd2, 'score']]) / \
                    np.sqrt(data.loc[:, idx[cnd1, 'SE']] ** 2 +
                            data.loc[:, idx[cnd2, 'SE']] ** 2)
                result_df.loc[:, idx[cnd1, cnd2, 'pvalue_raw']] = \
                    2 * stats.norm.sf(result_df.loc[:, idx[cnd1, cnd2, 'z']])

        self.store.put("/main/{}/scores_pvalues".format(label), result_df,
                       format="table")