anova_from_r

import pandas as pd
import numpy as np

class Anova:
  @classmethod
  def two_way(self, stack):
    """
    (対応のない)二元配置分散分析を行う。

    Anova.two_way(pd.DataFrame( 
      {'A': pd.Series(['a1', 'a1', 'a2', 'a2']),
       'B': pd.Series(['b1', 'b1', 'b2', 'b3']),
       'Y': pd.Series([  12,   14,   15,   24])})))
    ->
    Df    Sum Sq    Mean Sq    F value    Pr(>F) 
    A     1     42.25      42.25     21.125  0.136385
    B     1     40.50      40.50     20.250  0.139209
    Y     1      2.00       2.00        NaN       NaN

    """
    import pyper
    r = pyper.R()
    r.assign('stack', stack)
    A, B, y = stack.columns
    cmd = "res <- summary(aov(" + y + "~" + A + "+" + B + ", stack))"
    r(cmd)
    summary = r.get('res')[0]
    summary.index = stack.columns
    return summary