Question

我想从主成分回归中提取最低的MSEP。在下面的例子中，我想添加一行代码来说明＆＃34; [x]的最佳MSE是用[y]组件产生的，＆＃34;其中[x]是MSEP_object中的最小CV MSE（未调整的CV），[y]是产生该MSE的组件数。

library(MASS)
library(pls)
PCR_fit <- pcr(crim~., data=Boston, scale=TRUE, validation="CV")
MSEP_object <- MSEP(PCR_fit)

Answer 1

看了之后：

str(MSEP_object)
#------
List of 5
 $ val       : num [1:2, 1, 1:14] 74.1 74.1 51.8 51.8 51.8 ...
  ..- attr(*, "dimnames")=List of 3
  .. ..$ estimate: chr [1:2] "CV" "adjCV"
  .. ..$ response: chr "crim"
  .. ..$ model   : chr [1:14] "(Intercept)" "1 comps" "2 comps" "3 comps" ...
 $ type      : chr "MSEP"
 $ comps     : num [1:14] 0 1 2 3 4 5 6 7 8 9 ...
 $ cumulative: logi TRUE
 $ call      : language MSEP.mvr(object = PCR_fit)
 - attr(*, "class")= chr "mvrVal"

很容易看出“CV”-row是MSEP_object$val访问的数组的第一个维度。

> MSEP_object$val[1,1, ]
(Intercept)     1 comps     2 comps     3 comps     4 comps     5 comps     6 comps 
   74.13309    51.81121    51.77481    45.79611    45.60764    45.84421    46.42630 
    7 comps     8 comps     9 comps    10 comps    11 comps    12 comps    13 comps 
   46.32160    44.53877    44.83946    45.02798    44.91884    44.49455    43.51062 
> which.min(MSEP_object$val[1,1, ] )
13 comps 
      14

我不会称之为“最佳”，因为这意味着某种统计判断，但宁可暗示：

  paste( "Minimum MSE of ",  
         MSEP_object$val[1,1, ][ which.min(MSEP_object$val[1,1, ] )], 
         " was produced with ", 
         sub(" comps","", names(which.min(MSEP_object$val[1,1, ] ))), 
         " components")
 #[1] "Minimum MSE of  43.51061955445  was produced with  13  components"

从R中的主成分回归中提取最低MSEP

1 个答案: