library(dplyr)
 mtcars %>%
 group_by(vs) %>%
 do(tt=t.test(mpg~am, data=.)) %>%
 mutate(t=tt$statistic, p=tt$p.value)

I have performed t.test on mgp with respect to am for each vs. I am trying to extract the attributes of the t.test model and arrange them in a data frame. I am able to extract p-value and t-value or any one-length attribute.

Output:

Source: local data frame [2 x 4]
Groups: <by row>

vs         tt         t           p
1  0 <S3:htest> -2.579484 0.034481482
2  1 <S3:htest> -3.764735 0.004436935

Now, if I want to extract the mean group difference using the code below:

mtcars %>%
group_by(vs) %>%
do(tt=t.test(mpg~am, data=.)) %>%
mutate(t=tt$statistic, p=tt$p.value, delta=diff(tt$estimate))

or even when I just want to access the mean of group 0

     mutate(t=tt$statistic, p=tt$p.value, delta=tt$estimate[0])

R throws the error below:

  Error: object 'tt' not found

It seems tt is evaluated in a different environment where it is not defined when an expression or function is applied on it. I tried using funs too but no success. Could somebody please shed some light here?

We could extract it within the do without using any additional package.

mtcars %>%
    group_by(vs) %>%
    do({
      tt=t.test(mpg~am, data=.)
      data.frame(t=tt$statistic, p=tt$p.value,delta=diff(tt$estimate))
    })
#     vs         t           p    delta
#  (dbl)     (dbl)       (dbl)    (dbl)
#1     0 -2.579484 0.034481482 4.700000
#2     1 -3.764735 0.004436935 7.628571