R遍历组和列并将数据输出到帧中

Question

I'm trying to create a loop by group to save me having to test all variables in mctest for multicolinearity for a large dataset. As an example:

library(mctest)
library(AER)
library(dplyr)

iris <- datasets::iris
iris$stem <- sample(150, size = nrow(iris))
iris$lifespan <- sample(150, size=nrow(iris))
vif_results <- data.frame()

iris_setosa <- iris[which(iris$Species == "setosa"),]
i6 <- imcdiag(iris_setosa[,c(1:4)],iris_setosa[,6]) ## works
i7 <- imcdiag(iris_setosa[,c(1:4)],iris_setosa[,7]) ## works

##copy to df
i6_res <- data.frame(i6$idiags)
i6_res$group <- "setosa"
i6_res$Out <- "stem"

i7_res <- data.frame(i7$idiags)
i7_res$group <- "setosa"
i7_res$Out <- "life"

vif_results <- rbind(i6_res,i7_res)

##Rather than doing the above manually by Species and columns, write a loop to iterate through groups and columns 
for(col in 1:ncol(iris[,c(6:7)])){
  require(dplyr)
  iris1 <- iris %>%
  group_by(Species) %>% 
  do(imcdiag(iris[,c(1:4)],iris[,col]))
print(iris1)
}

f1 <- function(x){
  require(dplyr)
  iris1 <- iris %>%
    group_by(Species) %>% 
    do(imcdiag(iris[,c(1:4)],iris[,x]))
  return(f1)
}

for(col in 1:ncol(iris[,c(6:7)])){print(f1)}

The loops I've tried in various formats don't work. Thanks!

Edit

I've been playing with this and have been able to build the data frame in loop (I know that's not preferred). I still can't get the second loop to work to run the columns through the y-values so I've copied & edited to show what I'm trying to code with the 2nd y-value.

 for (i in seq_along(species)) { all <- subset(iris, iris$Species == species[i]) ## This is the part I wanted to create another loop for so that I don't have to list a lot of y-variables for bigger datasets y6 <- imcdiag(all[,1:4],all[,6]) ## This section works nicely to append all the vif values into a single table for easy analysis and export. vifs6 <- as.data.frame(y6$idiags) vifs6$Iteration <- i vifs6$Species <- species[[i]] vifs6$Output <- paste0(colnames(iris[6])) vif_results <- rbind(vif_results,vifs6) print(y6) y7 <- imcdiag(all[,1:4],all[,7]) vifs7 <- as.data.frame(y7$idiags) vifs7$Iteration <- i vifs7$Species <- species[[i]] vifs7$Output <- paste0(colnames(iris[7])) vif_results <- rbind(vif_results,vifs7) print(y7) } 

Any suggestions would be greatly appreciated!

Answer 1

With apply you can achieve your desired output, a data frame with results for all combinations of group and columns of interest.

We'll use the iris data frame with two extra columns, as described in your question

First, we create a data frame with all combinations or group and column of interest with expand.grid .

combinations <- expand.grid(names(iris[, 6:7]), unique(as.character(iris[["Species"]])))

Results in this:

      Var1       Var2
1     stem     setosa
2 lifespan     setosa
3     stem versicolor
4 lifespan versicolor
5     stem  virginica
6 lifespan  virginica

We have column names in the first column and group names in the second, so each row represents a pair of names.

With this, we can loop over each pair using apply .

results_list <- apply(combinations, 1, function(each_pair) {
  # An iris subset
  this_iris <- iris[which(iris$Species == each_pair[2]),]

  # Results for that iris subset and one of the columns
  result <- imcdiag(this_iris[,c(1:4)], this_iris[each_pair[1]])

  # Extract idiags as a data frame
  result_df <- as.data.frame(result$idiags)

  # Add columns with names for the group, column and variables
  result_df[["column"]] <- each_pair[1]
  result_df[["group"]] <- each_pair[2]
  result_df[["variable"]] <- rownames(result_df)
  result_df
})

We end up with a list of data frames. We can bind them together using do.call .

iris_idiags <- do.call(args = results_list, what = rbind)