which.min在一个向量的子集上

Question

我想得到一个向量的某个子集的最小值的索引,但是原始向量中的索引,而不是重新编号的子集.

至于现在,我一直在使用:

L = rnorm(20) # say this is the original vector
subset = runif(20)<0.3 # some conditions to extract the subset
ind_min = which.min(L[subset])
ind_sel = seq(L)[subset]
ind_min = ind_sel[ind_min]

但我想应该有更直接或更清洁的东西.我一直在考虑使用如下的技巧:

L_tmp = L
L_tmp[!subset] = Inf
ind_min = which.min(L_tmp)

这显然更有效:

> microbenchmark(method_1(), method_2(), unit = "relative")
Unit: relative
       expr      min       lq     mean   median       uq      max neval
 method_1() 3.699562 3.249635 3.119666 3.076819 2.928259 3.225849   100
 method_2() 1.000000 1.000000 1.000000 1.000000 1.000000 1.000000   100

但我对此并不满意,因为我猜应该还有别的东西.有什么建议吗？

Answer 1

你可以试试:

(seq(L))[subset][which.min(L[subset])]

这与您的第一个方法类似,但没有创建临时变量

20000长向量的基准结果L:

    method_cath<- function(){(seq(L))[subset][which.min(L[subset])]}
    method_FK_corr1 <- function(){min = min(L[subset])
                                  ind_min = intersect(which(L == min), seq(L)[subset])[1]
                                  return(ind_min)} 
    method_FK_corr2 <- function(){min = min(L[subset])
                                  ind_min = intersect(which(L == min), which(subset))[1]
                                  return(ind_min)} 
    method_1clm <- function(){ind_min = which.min(L[subset])
                              ind_sel = seq(L)[subset]
                              ind_min = ind_sel[ind_min]
                              return(ind_min)} 
    method_2clm <- function(){L_tmp = L
                              L_tmp[!subset] = Inf
                              ind_min = which.min(L_tmp)
                              return(ind_min)}

> microbenchmark(method_2clm(), method_cath(), method_1clm(), method_FK_corr2(), method_FK_corr1(), unit = "relative")
   # Unit: relative
   #               expr      min       lq     mean   median       uq       max neval cld
   #      method_2clm() 1.000000 1.000000 1.000000 1.000000 1.000000 1.0000000   100 a  
   #      method_cath() 1.312146 1.290370 1.282964 1.278178 1.282424 0.9191693   100  b 
   #      method_1clm() 1.295031 1.294642 1.303781 1.284630 1.279821 1.2977193   100  b 
   #  method_FK_corr2() 1.185821 1.166924 1.278030 1.155217 1.165738 4.9948007   100  b 
   #  method_FK_corr1() 1.683783 1.644797 1.746055 1.635293 1.636195 5.1616672   100   c

注意:我得到NA了@FedorenkoKristina原始函数的结果,我测试了2个可能的校正函数,现在所有函数都给出了相同的结果.