9

假设我有一个 data.table

set.seed(1) # to make the example reproducible
ex<-data.table(AAA=runif(100000),
               BBB=runif(100000),
               CCC=runif(100000),
               DDD=runif(100000),
               FLAG=c(rep(c("a","b","c","d","e"),200000)))

我想AAA每隔一列从列中减去，然后从BBB剩余的每一列（除 FLAG 除外）中减去，依此类推，以使输出看起来像......

ex[,list(AAA_BBB=AAA-BBB,
         AAA_CCC=AAA-CCC,
         AAA_DDD=AAA-DDD,
         BBB_CCC=BBB-CCC,
         BBB_DDD=BBB-DDD,
         CCC_DDD=CCC-DDD)]

是否有一个 data.table 语法可以在不知道有多少列或它们的名称是什么的情况下干净地做到这一点？

4

2 回答 2

5

Looping over the combinations within data.table:

comblist <- combn(names(ex)[-5],2,FUN=list)
res2 <- ex[,lapply(comblist,function(x) get(x[1])-get(x[2]))]

setnames(res2,names(res2),sapply(comblist,paste,collapse="_"))
于 2013-06-04T15:34:32.480 回答
5

combn与和的解决方案apply：

cc <- combn(colnames(ex)[1:4], 2)
apply(cc, 2, function(x)ex[[x[1]]]-ex[[x[2]]])

给出前 5 行：

             [,1]         [,2]       [,3]        [,4]        [,5]         [,6]
 [1,] -0.43500930 -0.520148152  0.1602265 -0.08513885  0.59523580  0.680374655
 [2,] -0.32964090 -0.153303302 -0.3807295  0.17633760 -0.05108855 -0.227426149
 [3,]  0.25991705 -0.079679566  0.2040904 -0.33959662 -0.05582670  0.283769917
 [4,]  0.35585252  0.153083047  0.2382553 -0.20276948 -0.11759719  0.085172292
 [5,] -0.67081018 -0.116543468 -0.3413471  0.55426671  0.32946305 -0.224803663

编辑

正如 Arun 建议的那样， combn 可以采用函数参数，因此更好的解决方案是

res <- combn(colnames(ex)[1:4], 2, function(x) ex[[x[1]]] - ex[[x[2]]])
colnames(res) <- combn(colnames(ex)[1:4], 2, paste, collapse="_")
as.data.table(res)

            AAA_BBB     AAA_CCC     AAA_DDD     BBB_CCC     BBB_DDD     CCC_DDD
      1: -0.4350093 -0.52014815  0.16022650 -0.08513885  0.59523580  0.68037465
      2: -0.3296409 -0.15330330 -0.38072945  0.17633760 -0.05108855 -0.22742615
      3:  0.2599171 -0.07967957  0.20409035 -0.33959662 -0.05582670  0.28376992
      4:  0.3558525  0.15308305  0.23825534 -0.20276948 -0.11759719  0.08517229
      5: -0.6708102 -0.11654347 -0.34134713  0.55426671  0.32946305 -0.22480366
     ---                                                                       
 999996: -0.8450458 -0.47951267 -0.30333929  0.36553310  0.54170648  0.17617338
 999997: -0.5778393 -0.01784418 -0.24353237  0.55999516  0.33430697 -0.22568819
 999998:  0.7127352  0.82554276  0.01258673  0.11280758 -0.70014846 -0.81295604
 999999: -0.6693544 -0.42335069 -0.81080852  0.24600375 -0.14145408 -0.38745783
1000000: -0.8511655 -0.23341818 -0.15830584  0.61774732  0.69285966  0.07511234
于 2013-06-04T15:13:38.913 回答