python计算显著性分组

包子甘蓝

于 2023-10-23 13:46:58 发布

阅读量86

点赞数

文章标签： python

本文链接：https://blog.csdn.net/weixin_43902858/article/details/133988213

版权

1、自定义函数

# 数据集 分组项 统计值
from statsmodels.stats.multicomp import pairwise_tukeyhsd
from statsmodels.stats.multicomp import MultiComparison

def pvaluegroup(data,groupcow,static):
    mc = MultiComparison(data[static], data[groupcow])
    result = mc.tukeyhsd()#计算各组间pvalue
    print(result.summary())
    #计算各组平均值，从高到低排序
    data_average=pd.DataFrame(data.groupby([groupcow])[static].mean())
    data_average=data_average.sort_values([static],ascending=False)
    data_average=data_average.reset_index(drop=False)
    data_average=data_average.astype(str)
    #修改格式
    dataresult=pd.DataFrame.from_records(result.summary())
    dataresult=dataresult.astype(str)
    # new_header = ["group1", "group2", "meandiff", "p-adj", "lower", "upper", "reject"]
    new_header = dataresult.iloc[0]
    dataresult.columns = new_header
    dataresult = dataresult[1:] # 删除第0行
    # dataresult=dataresult.astype(str)
    dataresult["p-adj"]=dataresult["p-adj"].astype(float)
    #设置显著性组
    data_average["显著性组"]=""
    p_level_abc_num=0#角标顺序
    for i in range(len(data_average[groupcow])):
        p_level=[]
        #角标字母
        p_level_abc=["a","b","c","d","e","f","g","h"]
    #     print(data_average.loc[i,groupcow])
        for j in range(i+1,len(data_average[groupcow])):
#             print(i,j,data_average.loc[i,groupcow],data_average.loc[j,groupcow])
            a=float(pd.concat([dataresult[(dataresult["group1"]==data_average.loc[i,groupcow])&(dataresult["group2"]==data_average.loc[j,groupcow])]["p-adj"],
                               dataresult[(dataresult["group1"]==data_average.loc[j,groupcow])&(dataresult["group2"]==data_average.loc[i,groupcow])]["p-adj"]]))
            if a<0.05 :#p=0.05
                p_level.append(data_average.loc[j,groupcow])
                #追加显著性组
                data_average.loc[j,"显著性组"]=str(data_average.loc[j,"显著性组"])+p_level_abc[p_level_abc_num]
#                 print(p_level)
        if p_level != []:
            data_average.loc[i,"显著性组"]=str(data_average.loc[i,"显著性组"])+p_level_abc[i]
            p_level_abc_num+=1  #角标移至下一位
        
    #     data_average.loc[i,"显著性组"]=p_level
    data_average=data_average.sort_values([groupcow]) #按组别升序
    return data_average

数据集、数据分组列、数据值列，分组列可以为数字或字母，不同水平需合成一列，如：公母两个性别、大白长白两个品种，需合并成一列，分别对应大白公、大白母、长白公、长白母
输出结果1
输出结果2

2、匹配数据

df_out_group=pd.merge(a_30_pgrouop,a_100_pgrouop,left_on="品种性别",right_on="品种性别")
df_out_group=pd.merge(df_out_group,a_120_pgrouop,left_on="品种性别",right_on="品种性别")
df_out_group['品种性别']=df_out_group['品种性别'].replace({"1.0":'杜洛克公', "2.0":'杜洛克母', "3.0":'长白公', "4.0":'长白母', "5.0":'大白公', "6.0":'大白母'})
df_out_group