Running multiple T-Test on variables with groupings in R (not using rstatix)

Question

I've a data frame containing different items (and it's cost) and also it's subsequent groupings. I would like to run a T-Test for each item based on their groupings to see if their mean differs. Anybody knows how to do this in R without using the rstatix package? If possible, done in base R using lapply or looping. Tidyr and dplyr is fine.

A sample of the dataframe is as follow:

df = structure(list(Item = structure(c(1L, 1L, 1L, 1L, 2L, 2L, 2L, 
2L, 3L, 3L, 3L, 3L, 3L, 4L, 4L, 4L, 4L), .Label = c("Book A", 
"Book B", "Book C", "Book D"), class = "factor"), Cost = c(7L, 
9L, 6L, 7L, 4L, 6L, 5L, 3L, 5L, 4L, 7L, 2L, 2L, 4L, 2L, 9L, 4L
), Grouping = structure(c(1L, 2L, 1L, 2L, 1L, 2L, 1L, 1L, 2L, 
1L, 1L, 2L, 2L, 1L, 2L, 2L, 1L), .Label = c("A", "B"), class = "factor")), class = "data.frame", row.names = c(NA, 
-17L))

Item	Cost	Grouping
Book A	7	A
Book A	9	B
Book A	6	A
Book A	7	B
Book B	4	A
Book B	6	B
Book B	5	A
Book B	3	A
Book C	5	B
Book C	4	A
Book C	7	A
Book C	2	B
Book C	2	B
Book D	4	A
Book D	2	B
Book D	9	B
Book D	4	A

The output should be a simple table (or any similar table) as follows

Item	P-Value (H0: Mean of group A = Mean of group B)
Book A	xxx
Book B	xxx
Book C	xxx
Book D	xxx

Using the rstatix package, the code will be (credits: Quinten)

library(dplyr)
library(rstatix)
df %>% 
  group_by(Item) %>%
  t_test(Cost ~ Grouping)

I would like to achieve the same output but without using rstatix package as I often encounter issues with the broom package (dependent package of rstatix). Base package would be fine as I code with my phone sometimes.

Thank you!

Answer 1

The error relates to the number of observations in 'Grouping'. There is a case of having 1 observation. With base R, we can do this as

lapply(split(df, df$Item), function(x) if(any(table(x$Grouping) < 2)) 
      NA else t.test(Cost ~ Grouping, data = x))

-output

$`Book A`

    Welch Two Sample t-test

data:  Cost by Grouping
t = -1.3416, df = 1.4706, p-value = 0.3499
alternative hypothesis: true difference in means between group A and group B is not equal to 0
95 percent confidence interval:
 -8.418523  5.418523
sample estimates:
mean in group A mean in group B 
            6.5             8.0 


$`Book B`
[1] NA

$`Book C`

    Welch Two Sample t-test

data:  Cost by Grouping
t = 1.3868, df = 1.8989, p-value = 0.3059
alternative hypothesis: true difference in means between group A and group B is not equal to 0
95 percent confidence interval:
 -5.666332 10.666332
sample estimates:
mean in group A mean in group B 
            5.5             3.0 


$`Book D`

    Welch Two Sample t-test

data:  Cost by Grouping
t = -0.42857, df = 1, p-value = 0.7422
alternative hypothesis: true difference in means between group A and group B is not equal to 0
95 percent confidence interval:
 -45.97172  42.97172
sample estimates:
mean in group A mean in group B 
            4.0             5.5 

Or getting the pvalue

stack(lapply(split(df, df$Item), function(x) if(any(table(x$Grouping) < 2)) 
      NA else t.test(Cost ~ Grouping, data = x)$p.value))[2:1]
  ind    values
1 Book A 0.3498856
2 Book B        NA
3 Book C 0.3058987
4 Book D 0.7422379

---

The same approach can be done with dplyr

library(dplyr)
df %>% 
  add_count(Item, Grouping) %>%
  group_by(Item) %>%
   summarise(out = list(if(any(n < 2)) NA else t.test(Cost ~ Grouping)))

-output

# A tibble: 4 × 2
  Item   out      
  <fct>  <list>   
1 Book A <htest>  
2 Book B <lgl [1]>
3 Book C <htest>  
4 Book D <htest>  

If it needs only the pvalue

df %>% 
  add_count(Item, Grouping) %>%
  group_by(Item) %>%
   summarise(out = if(any(n < 2)) NA_real_ else t.test(Cost ~ Grouping)$p.value)
# A tibble: 4 × 2
  Item      out
  <fct>   <dbl>
1 Book A  0.350
2 Book B NA    
3 Book C  0.306
4 Book D  0.742