Question

我有一个数据，我试图使用double for循环聚合。基本上，我想通过每个TOP_LEVEL_CATEGORY计算每个MEM_ID的购买周期。数据如下所示，

MEM_ID  ORDER_DEL_DATE  TOP_LEVEL_CATEGORY
999984  2016-01-07         household
999984  2016-02-03         household
999980  2015-12-16         household
999980  2016-01-03         household
999980  2016-01-05         household
999980  2016-02-14         household
999984  2016-01-07         personal-care
999980  2016-01-03         personal-care
999980  2016-01-30         personal-care

代码

PC_test <- NA
for(i in unique(test$MEM_ID)){
  for(j in unique(test$TOP_LEVEL_CATEGORY)){
    PC_test[c(i,j)] <- data.frame(c(MEM_ID=i,CATEGORY=j,ifelse(nrow(test[test$MEM_ID==i & test$TOP_LEVEL_CATEGORY==j,])<=2, 
                 max(test[test$MEM_ID==i & test$TOP_LEVEL_CATEGORY==j,"ORDER_DEL_DATE"])-min(test[test$MEM_ID==i & test$TOP_LEVEL_CATEGORY==j,"ORDER_DEL_DATE"]),
                  max(test[test$MEM_ID==i & test$TOP_LEVEL_CATEGORY==j,"ORDER_DEL_DATE"])-maxN(test[test$MEM_ID==i & test$TOP_LEVEL_CATEGORY==j,"ORDER_DEL_DATE"]))))
  }
}

注意：maxN函数给出了第二大函数。

抛出一个无关的输出，如下所示，

NA. X999984    household    personal.care   X999980
NA       999984      999980      999980         999980
NA   personal-care  household   personal-care   personal-care
NA    0               40           27                27

我希望以下格式输出

 MEM_ID       TOP_LEVEL_CATEGORY   PC_test
 999984           household          27
 999984           personal-care      0
 999980           household          40
 999980           personal-care      27

非常感谢帮助。在此先感谢!!!

Answer 1

我想你想要这样的东西

require(data.table)

setDT(df1)

#  calculate the min and max date for each MEM_ID/TOP_LEVEL_CATEGORY pair, then find the difference for PC_test
df1[,  .(max_date=max(ORDER_DEL_DATE), min_date=min(ORDER_DEL_DATE)),
  keyby=.(MEM_ID,TOP_LEVEL_CATEGORY)][, .(MEM_ID, TOP_LEVEL_CATEGORY, PC_test = max_date - min_date)]


   MEM_ID TOP_LEVEL_CATEGORY PC_test
1: 999980          household 60 days
2: 999980      personal-care 27 days
3: 999984          household 27 days
4: 999984      personal-care  0 days

根据你的说法，这是我对你如何计算PC_test而不完全分解你的最小和最大公式的有根据的猜测。

R中的双循环给出无关的输出

1 个答案: