Consider a dataset Data which has several factor and several numerical continuous variables. Some of these variables, let's say slice_by_1 (with classes "Male", "Female") and slice_by_2 (with classes "Sad", "Neutral", "Happy"), are used to 'slice' data into subsets. For every subset Kruskal-Wallis test should be run on variables length, preasure,pulse each grouped by the other factor variable called compare_by. Is there a quick way in R to accomplish this task and put calculated p values to a matrix?
I used dplyr package to prepare data.
Sample dataset:
library(dplyr)
set.seed(123)
Data <- tbl_df(
data.frame(
slice_by_1 = as.factor(rep(c("Male", "Female"), times = 120)),
slice_by_2 = as.factor(rep(c("Happy", "Neutral", "Sad"), each = 80)),
compare_by = as.factor(rep(c("blue", "green", "brown"), times = 80)),
length = c(sample(1:10, 120, replace=T), sample(5:12, 120, replace=T)),
pulse = runif(240, 60, 120),
preasure = c(rnorm(80,1,2),rnorm(80,1,2.1),rnorm(80,1,3))
)
) %>%
group_by(slice_by_1, slice_by_2)
Let's look at data:
Source: local data frame [240 x 6]
Groups: slice_by_1, slice_by_2
slice_by_1 slice_by_2 compare_by length pulse preasure
1 Male Happy blue 10 69.23376 0.508694601
2 Female Happy green 1 68.57866 -1.155632020
3 Male Happy brown 8 112.72132 0.007031799
4 Female Happy blue 3 116.61283 0.383769524
5 Male Happy green 7 110.06851 -0.717791526
6 Female Happy brown 8 117.62481 2.938658488
7 Male Happy blue 9 105.59749 0.735831389
8 Female Happy green 2 83.44101 3.881268679
9 Male Happy brown 5 101.48334 0.025572561
10 Female Happy blue 10 62.87331 -0.715108893
.. ... ... ... ... ... ...
An example of desired output:
Data_subsets length preasure pulse
1 Male_Happy <p-value> <p-value> <p-value>
2 Female_Happy <p-value> <p-value> <p-value>
3 Male_Neutral <p-value> <p-value> <p-value>
4 Female_Neutral <p-value> <p-value> <p-value>
5 Male_Sad <p-value> <p-value> <p-value>
6 Female_Sad <p-value> <p-value> <p-value>
We could use Map within do for doing the multiple column kruskal.test and then use unite from library(tidyr) to join the 'slice_by_1' and 'slice_by_2' columns to a single column 'Data_subsets'.
library(dplyr)
library(tidyr)
nm1 <- names(Data)[4:6]
f1 <- function(x,y) kruskal.test(x~y)$p.value
Data %>%
do({data.frame(Map(f1, .[nm1], list(.$compare_by)))}) %>%
unite(Data_subsets, slice_by_1, slice_by_2, sep="_")
# Data_subsets length pulse preasure
#1 Female_Happy 0.4369918 0.8767561 0.1937327
#2 Female_Neutral 0.3750688 0.2858796 0.8588069
#3 Female_Sad 0.7958502 0.5801208 0.6274940
#4 Male_Happy 0.3099704 0.3796494 0.6929493
#5 Male_Neutral 0.4953853 0.2418708 0.2986860
#6 Male_Sad 0.7159970 0.5686672 0.8528201
Or we can do this using data.table. We convert the 'data.frame' to 'data.table' (setDT(Data)), create grouping variable ('Data_subsets') by pasteing the 'slice_by_1' and 'slice_by_2' columns, then we subset the columns of the dataset and pass that as input to Map, do the krusal.test and extract the p.value.
library(data.table)
setDT(Data)[, Map(f1, .SD[, nm1, with=FALSE], list(compare_by)) ,
by = .(Data_subsets= paste(slice_by_1, slice_by_2, sep='_'))]
# Data_subsets length pulse preasure
#1: Male_Happy 0.3099704 0.3796494 0.6929493
#2: Female_Happy 0.4369918 0.8767561 0.1937327
#3: Male_Neutral 0.4953853 0.2418708 0.2986860
#4: Female_Neutral 0.3750688 0.2858796 0.8588069
#5: Male_Sad 0.7159970 0.5686672 0.8528201
#6: Female_Sad 0.7958502 0.5801208 0.6274940