R - dplyr bootstrap several variables

Question

I am trying out bootstraping with dplyr and I am stuck with a simple line of code.

Using the function bootstrap, I found out that it is possible to do

library(dplyr) library(broom) mtcars %>% bootstrap(10) %>% do(tidy(sample(.$cyl, 2)))

to get a nice straightforward output

 replicate x (int) (dbl) 1 1 6 2 1 8 3 2 6 4 2 8 ...

However, it would be nice to get more variables (columns), but I can't figure it how.

I thought something like

mtcars %>% bootstrap(10) %>% do(tidy(sample(., 2)))

or

mtcars %>% bootstrap(10) %>% do(tidy(sample_n(2)))

would work but it doesn't.

Any clue how I can subset several variables ?

Imagine I want to get mpg, cyl and disp to get something like (output)

 replicate cyl mpg disp (int) (dbl) 1 1 6 21 ... 2 1 4 22 ... 3 2 6 ... 4 2 8 ... ...

(I am randomly choosing two cases sample = 2 and I repeat this routine (bootstrap) 10 times).

Is your goal just to get a bunch of random samples w/ replacement for each column in the original data frame? What is the model you are ultimately trying to fit? There are some more examples on bootstrapping with broom in the bootstrapping vignette that may be helpful. — Keith Hughitt
– Keith Hughitt, Commented Aug 16, 2016 at 10:51
@Keith, thanks I'll look at it. It is not to fit a model, just data manipulation. Whoever voted down, can I have an explanation ? thanks — giac
– giac, Commented Aug 16, 2016 at 16:23

renato vitolo · Accepted Answer · 2016-08-16 11:36:34Z

Using

set.seed(123) sapply(mtcars, function(v) sample(v,2))

you can sample 2 values from each column of mtcars, where however the columns are sampled independently of each other (not sure that this is what you want and/or that it makes sense). Therefore a solution using broom might be:

mtcars %>% bootstrap(10) %>% do(tidy(sapply(., function(v) sample(v,2))))

If, on the other hand, preserving the relations between the columns is important, you could use something like

do.call("rbind",lapply(1:10, function(dum) mtcars[sample.int(nrow(mtcars), 2), ]))

Collectives™ on Stack Overflow

R - dplyr bootstrap several variables

1 Answer 1

Comments

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Related