0

I wanted to run t-tests in a dataset that have two factors: A and B. I've got this working but without removing the outliers.

My idea is to remove values using the 1.5*IQR criteria. I've could not figure out how to do this following the dyplyr way (?).

Here is what I have:

wallSize %>%
  select(Time,  A, B) %>%
  gather(key = variable, value = value, -A, -B) %>% 
  group_by(A, B, variable) %>%
  summarise(value = list(value)) %>% 
  spread(A, value) %>%
  group_by(variable) %>%
  mutate(p_value = t.test(unlist(True), unlist(False), paired=TRUE)$p.value,
         t_value = t.test(unlist(True), unlist(False), paired=TRUE)$statistic)))

I think I should do the outlier removal after the spread for each of the 6 lists individually but I can't figure it out how... Any suggestions from the R masters?

Cheers

EDIT: Sample data head of the frame before grouping:

  Display Change Comp       TargetType TotalTime SelectionTime Score
  <chr>   <chr>  <chr>                <int>     <dbl>         <dbl> <int>
1 Wall    Shape  False                    1     62.2          53.7      4
2 Wall    Shape  False                    2     14.1          12.6      5
3 Wall    Shape  True                     0     26.3          23.0      5
4 Wall    Shape  True                     0     20.3          14.7      5
5 Wall    Shape  True                     1     23.3          21.6      5
6 Wall    Shape  False                    2      6.55          5.17     5

after grouping:

  TargetType variable      False      True      
       <int> <chr>         <list>     <list>    
1          0 SelectionTime <dbl [28]> <dbl [28]>
2          1 SelectionTime <dbl [28]> <dbl [28]>
3          2 SelectionTime <dbl [28]> <dbl [28]>
Sal-laS
  • 11,016
  • 25
  • 99
  • 169
Juliano Franz
  • 61
  • 8
  • Could you please provide an example of your data or data like it? https://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example – Jon Spring Sep 13 '18 at 05:08
  • Hi Jon, I've edit the question with the head of the df and the output of the grouped data. Thanks for trying to help. People are quick to downvote =/ – Juliano Franz Sep 13 '18 at 10:46

0 Answers0