How to create a data frame from R output

Question

I'm trying to create a dataset from the output of multiple operations. But I don't know how to automate this. The replicate function would be nice but there's multiple operations to perform to get the single new data points ie, adjusted R squared & F-statistic.

R code:

#make dataframe with random data
A<-as.integer(round(runif(20, min=1, max=10)))
dim(A) <- c(10,2)
A<-as.data.frame(A)
#extract F-statistic
summary(lm(formula=V1~V2,data=A))$fstatistic[1]
#extract adjusted R squared
summary(lm(formula=V1~V2,data=A))$adj.r.squared
#repeat 100 times and make a dataframe of the unique extracted output, e.g. 2 columns 100 rows
??????????????

Answer 1

Applying the linear model over 5 data frames...

With replicate , it would be something like

> replicate(5, {
      A <- data.frame(rnorm(5), rexp(5))
      m <- lm(formula = A[,1] ~ A[,2], data = A)
      c(f = summary(m)$fstatistic[1], adjR = summary(m)$adj.r.squared)
  })
##               [,1]      [,2]       [,3]      [,4]        [,5]
## f.value  0.4337426 1.3524681 1.17570087 3.8537837  0.04583862
## adjR    -0.1649097 0.0809812 0.04207698 0.4163808 -0.31326721

And you can wrap this with t() to get the long form matrix.

You could also use the ever-popular do.call(rbind, lapply(...)) method,

> do.call(rbind, lapply(seq(5), function(x){
      A <- data.frame(rnorm(5), rexp(5))
      m <- lm(formula = A[,1] ~ A[,2], data = A)
      c(f = summary(m)$fstatistic[1], adjR = summary(m)$adj.r.squared)
  }))
##          f.value        adjR
## [1,]   1.9820243  0.19711351
## [2,]  21.6698543  0.83785879
## [3,]   4.4484639  0.46297652
## [4,]   0.9084373 -0.02342693
## [5,]   0.0388510 -0.31628698

You can also use sapply ,

> sapply(seq(5), function(x){
      A <- data.frame(rnorm(5), rexp(5))
      m <- lm(formula = A[,1] ~ A[,2], data = A)
      c(f = summary(m)$fstatistic[1], adjR = summary(m)$adj.r.squared)
  })
##                    [,1]       [,2]          [,3]       [,4]        [,5]
## f.value      0.07245221  0.2076504  0.0003488657 58.5524139  0.92170453
## adjR        -0.30189169 -0.2470187 -0.3331783000  0.9350147 -0.01996465

Keep in mind that these all return a matrix , so an as.data.frame wrapper may be appropriate if you would like a data.frame result.

Answer 2

Just wrap it up in a for loop.

df <- as.data.frame(matrix(0, 100, 2))

for (i in 1:100){
 A<-as.integer(round(runif(20, min=1, max=10)))
 dim(A) <- c(10,2)
 A<-as.data.frame(A)
 #extract F-statistic
 df[i, 1] <- summary(lm(formula=V1~V2,data=A))$fstatistic[1]
 #extract adjusted R squared
 df[i, 2] <- summary(lm(formula=V1~V2,data=A))$adj.r.squared
}

Viola.

Answer 3

The replicate function will work fine. First, write a function to do one iteration of your simulation.

one.sim <- function() {
    A <- matrix(as.integer(runif(20, min=1, max=10)), nrow=10)
    A <- as.data.frame(A)
    m1.summary <- summary(lm(V1 ~ V2, data=A))
    return(c(fstatistic=unname(m1.summary$fstatistic[1]), 
             adj.r.squared=m1.summary$adj.r.squared))
}

Then use this function in replicate:

results <- t(replicate(100, one.sim()))