Question

我已经知道如何通过数据表中的名称和数字来调用列。

我有兴趣以更灵活的方式调用和执行转化。例如...假设我正在使用灵活创建的数据库（J可以是任何东西）。

  J = 3
  set.seed(1)
  temp0 = do.call(CJ, replicate(J, rnorm(10,0,1), simplify = FALSE))
  tempp = do.call(CJ, replicate(J, seq(0,1,length.out=10), simplify = FALSE))
  setnames(temp0, paste0('beta', 1:J))
  setnames(tempp, paste0('p', 1:J))

  temp0[,pc:=rnorm(10,-1,.1)]
  temp3 = cbind(temp0,tempp)

提供输出

> temp3
           beta1     beta2       beta3         pc p1 p2        p3
   1: -0.8356286 -2.214700 -1.98935170 -0.8641320  0  0 0.0000000
   2: -0.8356286 -2.214700 -1.47075238 -1.0102788  0  0 0.1111111
   3: -0.8356286 -2.214700 -0.47815006 -0.9612328  0  0 0.2222222
   4: -0.8356286 -2.214700 -0.15579551 -1.0053805  0  0 0.3333333
   5: -0.8356286 -2.214700 -0.05612874 -1.1377060  0  0 0.4444444
  ---                                                            
 996:  1.5952808  1.511781  0.07456498 -1.0414995  1  1 0.5555556
 997:  1.5952808  1.511781  0.41794156 -1.0394290  1  1 0.6666667
 998:  1.5952808  1.511781  0.61982575 -1.0059313  1  1 0.7777778
 999:  1.5952808  1.511781  0.78213630 -0.8899975  1  1 0.8888889
1000:  1.5952808  1.511781  0.91897737 -0.9236824  1  1 1.0000000

然后我想创建一个称为denom和pi {j}（每个J代表一个变量）的变量，例如，它遵循以下公式。

temp3[,denom:=(1+exp(beta1+pc*p1) + exp(beta2+pc*p2) + exp(beta3+pc*p3))]
temp3[,pi1:=(exp(beta1+pc*p1)/denom)]
temp3[,pi2:=(exp(beta2+pc*p2)/denom)]
temp3[,pi3:=(exp(beta3+pc*p3)/denom)]

提供输出：

> temp3
           beta1     beta2       beta3         pc p1 p2        p3    denom       pi1        pi2        pi3
   1: -0.8356286 -2.214700 -1.98935170 -0.8641320  0  0 0.0000000 1.679572 0.2581621 0.06500839 0.08143983
   2: -0.8356286 -2.214700 -1.47075238 -1.0102788  0  0 0.1111111 1.748145 0.2480354 0.06245836 0.11747135
   3: -0.8356286 -2.214700 -0.47815006 -0.9612328  0  0 0.2222222 2.043484 0.2121876 0.05343145 0.24502052
   4: -0.8356286 -2.214700 -0.15579551 -1.0053805  0  0 0.3333333 2.154850 0.2012214 0.05067002 0.28403911
   5: -0.8356286 -2.214700 -0.05612874 -1.1377060  0  0 0.4444444 2.112984 0.2052083 0.05167398 0.26985329
  ---                                                                                                     
 996:  1.5952808  1.511781  0.07456498 -1.0414995  1  1 0.5555556 4.944346 0.3518806 0.32369194 0.12217626
 997:  1.5952808  1.511781  0.41794156 -1.0394290  1  1 0.6666667 5.106751 0.3413962 0.31404743 0.14873716
 998:  1.5952808  1.511781  0.61982575 -1.0059313  1  1 0.7777778 5.311170 0.3394384 0.31224650 0.16003264
 999:  1.5952808  1.511781  0.78213630 -0.8899975  1  1 0.8888889 5.877735 0.3444218 0.31683070 0.16861387
1000:  1.5952808  1.511781  0.91897737 -0.9236824  1  1 1.0000000 5.753231 0.3402198 0.31296531 0.17299948

但是，这是硬编码为3的，我该如何对任意J执行此操作？这个问题非常困难，因为它似乎涉及调用列号并同时进行转换。

Answer 1

远非如此，但是您可以使用Map来循环遍历使用mget选择并分配给"pi1/2/3"的每个不同输入。我已经稍微改变了函数的顺序，以首先计算基本的重复部分，然后是分母，然后除以分母：

J = 3
temp3[,
     (paste0("pi",seq(J))) := Map(
        function(b,pc,p) exp(b + pc * p),
        mget(paste0("beta", seq(J))),
        .(pc),
        mget(paste0("p", seq(J)))
    )
]
temp3[, denom := 1 + Reduce(`+`, mget(paste0("pi", seq(J))) ) ]
temp3[, (paste0("pi",seq(J))) := lapply(mget(paste0("pi", seq(J))), `/`, denom) ]

似乎给出相同的结果：

temp3[, .(pi1,pi2,pi3,denom)]
#            pi1        pi2        pi3    denom
#   1: 0.2581621 0.06500839 0.08143983 1.679572
#   2: 0.2480354 0.06245836 0.11747135 1.748145
#   3: 0.2121876 0.05343145 0.24502052 2.043484
#   4: 0.2012214 0.05067002 0.28403911 2.154850
#   5: 0.2052083 0.05167398 0.26985329 2.112984
#  ---                                         
# 996: 0.3518806 0.32369194 0.12217626 4.944346
# 997: 0.3413962 0.31404743 0.14873716 5.106751
# 998: 0.3394384 0.31224650 0.16003264 5.311170
# 999: 0.3444218 0.31683070 0.16861387 5.877735
#1000: 0.3402198 0.31296531 0.17299948 5.753231

数据表中的动态列调用和转换

1 个答案: