Apply function per row in R

user_13642271 0
Viewed 80

I have a dataset to which I am trying to apply shapiro.test.

test_data <- tibble(
  gene = rep(c(LETTERS[1:5]), times = 2, each = 5),
  treatment = c(rep('control', times = 25) , rep('treatment', times = 25)),
  day = rep(c(1:2), times = 5, each = 5),
  data = rnorm(50, mean = 25, sd = 5)
)

# A tibble: 50 x 4
   gene  treatment   day  data
   <chr> <chr>     <int> <dbl>
 1 A     control       1  28.8
 2 A     control       1  22.4
 3 A     control       1  24.8
 4 A     control       1  20.1
 5 A     control       1  15.6
 6 B     control       2  26.5
 7 B     control       2  26.2
 8 B     control       2  25.3
 9 B     control       2  21.4
10 B     control       2  35.0
# … with 40 more rows

I created a function to run the test per gene, treatment and day:

normality_test <- function(x, y, z){
 with(test_data, shapiro.test(data[gene == x & treatment == y & day == z]))
}

So, if I run normality_test('A', 'control', '1') it will test gene A in control on day 1.

    Shapiro-Wilk normality test

data:  data[gene == x & treatment == y & day == z]
W = 0.99935, p-value = 0.9998

However, I want the function to loop through all combinations of gene/treatment/day and output each normality test individually, but have not been able to figure it out.
I have been able to create a loop that outputs each row as a separate tibble, but have not been successful in separating each element in the row to add to the normality_test function.
I also experimented with map and lmap, but to no avail.
Thank you.

purrr r for-loop
2 Answers
aggregate(test_data$data, 
          list(test_data$gene, test_data$treatment, test_data$day),
          FUN = shapiro.test, 
          simplify = FALSE)

leading to:

   Group.1   Group.2 Group.3                                                                          x
1        A   control       1   0.88967912801297, 0.355492282929607, Shapiro-Wilk normality test, X[[i]]
2        C   control       1  0.872622601121686, 0.277206156703845, Shapiro-Wilk normality test, X[[i]]
3        E   control       1  0.886337216320072, 0.339020498506711, Shapiro-Wilk normality test, X[[i]]
4        B treatment       1  0.902918723585913, 0.426227078990388, Shapiro-Wilk normality test, X[[i]]
5        D treatment       1  0.850079117181635, 0.194768724952993, Shapiro-Wilk normality test, X[[i]]
6        B   control       2 0.810238329506977, 0.0979423168617567, Shapiro-Wilk normality test, X[[i]]
7        D   control       2  0.965126339172019, 0.843147936412715, Shapiro-Wilk normality test, X[[i]]
8        A treatment       2  0.933316692276928, 0.619155973369246, Shapiro-Wilk normality test, X[[i]]
9        C treatment       2 0.771672137756979, 0.0466697848359151, Shapiro-Wilk normality test, X[[i]]
10       E treatment       2   0.91332590966644, 0.487832813430558, Shapiro-Wilk normality test, X[[i]]
user_65031410 
library(dplyr)

test_data %>% 
  nest_by(gene, treatment, day, .key = "nested") %>% 
  mutate(sw = list(shapiro.test(nested$data))) %>% 
  pull(sw)

Output

You have 10 unique combinations of gene, treatment, and day in your example data.

Showing the first few elements:

[[1]]

    Shapiro-Wilk normality test

data:  nested$data
W = 0.88041, p-value = 0.3112


[[2]]

    Shapiro-Wilk normality test

data:  nested$data
W = 0.96533, p-value = 0.8445

If you don't pipe to pull then your output will be:

   gene  treatment   day             nested sw     
   <chr> <chr>     <int> <list<tibble[,1]>> <list> 
 1 A     control       1            [5 x 1] <htest>
 2 A     treatment     2            [5 x 1] <htest>
 3 B     control       2            [5 x 1] <htest>
 4 B     treatment     1            [5 x 1] <htest>
 5 C     control       1            [5 x 1] <htest>
 6 C     treatment     2            [5 x 1] <htest>
 7 D     control       2            [5 x 1] <htest>
 8 D     treatment     1            [5 x 1] <htest>
 9 E     control       1            [5 x 1] <htest>
10 E     treatment     2            [5 x 1] <htest>

Where column sw has the results of the test.

Or for just the p-values:

test_data %>% 
  nest_by(gene, treatment, day, .key = "nested") %>% 
  mutate(sw = shapiro.test(nested$data)$p.value)

Output

   gene  treatment   day             nested    sw
   <chr> <chr>     <int> <list<tibble[,1]>> <dbl>
 1 A     control       1            [5 x 1] 0.311
 2 A     treatment     2            [5 x 1] 0.845
 3 B     control       2            [5 x 1] 0.408
 4 B     treatment     1            [5 x 1] 0.204
 5 C     control       1            [5 x 1] 0.435
 6 C     treatment     2            [5 x 1] 0.316
 7 D     control       2            [5 x 1] 0.143
 8 D     treatment     1            [5 x 1] 0.236
 9 E     control       1            [5 x 1] 0.695
10 E     treatment     2            [5 x 1] 0.658
user_63824340
Related