Question

我想使用R的Fisher's exact test functionality（特别是MC模拟功能）和Python的接口。我试图使用rpy2来做到这一点，但这比我想象的要困难。

我可以使用以下代码获得Fisher测试方法的界面：

import rpy2.robjects as robjects
fisher = robjects.r['fisher.test']

但是，如何将2xN矩阵传递给函数并检索p值？

Answer 1

考虑导入R的stats包并将Fisher Test作为Python函数运行。请注意，result对象是<class 'rpy2.robjects.vectors.ListVector'>，因此必须转换为Python字典，如下所示。

import rpy2
from rpy2.robjects.numpy2ri import numpy2ri
from rpy2.robjects.packages import importr
import numpy as np

cont = np.reshape(np.arange(0,4), (2,2))

statspackage = importr('stats',  robject_translations={'format_perc': '_format_perc'})    
result = statspackage.fisher_test(numpy2ri(cont), simulate_p_value = True, B = 100)

# DEPRECATED CONVERSION
import pandas.rpy.common as com
pyresultdict = com.convert_robj(result)

for k, v in  pyresultdict.items():
    print(k, v)

# data.name ['structure(c(0L, 2L, 1L, 3L), .Dim = c(2L, 2L))']
# p.value [1.0]
# estimate odds ratio    0.0
# dtype: float64
# null.value odds ratio    1.0
# dtype: float64
# conf.int [0.0, 77.90626902008512]
# alternative ['two.sided']
# method ["Fisher's Exact Test for Count Data"]

另一方面，您可能会收到有关com.convert_to_r_dataframe和com.convert_robj(rdf)弃用的警告，应根据建议here替换为pandas2ri.pandas2ri()和pandas2ri。但是，我的转换不适用于 ListVector 对象。理想情况下，上述转换将替换为以下内容：

# CURRENT CONVERSION
from rpy2.robjects import pandas2ri
pandas2ri.activate()
pyresultdict = pandas2ri.ri2py(result)

for k, v in  pyresultdict.items():
    print(k, v)

Answer 2

以下是一种方法：

import rpy2.robjects as robjects
from rpy2.robjects import r
from rpy2.robjects.numpy2ri import numpy2ri
from rpy2.robjects.packages import importr
import numpy as np

cont = np.reshape(np.arange(0,4), (2,2))
print cont

r_cont = numpy2ri(cont)
r.assign("cont", r_cont)
r("res <- fisher.test(cont, simulate.p.value = TRUE, B = 100)")
r_result = r("res")
p_value = r_result[0][0]
print r_result
print p_value

将数据从Python发送到R以使用rpy2执行统计测试

2 个答案: