I'm having a problem I can't figure out... Basically I want to generate mean, SD, and N per group for a number of variables. My data looks like this:
dataSet <- data.frame(study_id=c(1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4),
Timepoint=c(1,6,12,18,1,6,12,18,1,6,12,18,1,6,12,18),
Secretor=c(0,0,0,0,1,1,1,1,0,0,0,0,1,1,1,1),
Gene1=runif(16, min=0, max=100),
Gene2=runif(16, min=0, max=100),
Gene3=runif(16, min=0, max=100),
Gene4=runif(16, min=0, max=100))
Then I group it...
library(tidyverse)
grouped_dataSet <- dataSet %>%
group_by(Secretor, Timepoint)
When I run the following line of code, I get what I want:
summarise(grouped_dataSet, mean = mean(Gene1, na.rm=T), sd = sd(Gene1, na.rm=T), n = n())
Output:
# A tibble: 8 x 5
# Groups: Secretor [2]
Secretor Timepoint mean sd n
<dbl> <dbl> <dbl> <dbl> <int>
1 0 1 21.8 18.6 2
2 0 6 34.8 33.2 2
3 0 12 43.1 4.34 2
4 0 18 72.6 38.0 2
5 1 1 13.3 15.3 2
6 1 6 41.2 22.8 2
7 1 12 44.9 25.7 2
8 1 18 37.0 8.49 2
However, when I write this same line of code as a function (which I'm intending to then map onto many columns using tidyverse's purrr package), it doesn't work, instead returning "NA" for everything except the n column:
summary_function <- function(x) {
summary <- summarise(grouped_dataSet, mean = mean(x, na.rm=T), sd = sd(x, na.rm=T), n = n())
return(summary)
}
summary_function("Gene1")
Output:
# A tibble: 8 x 5
# Groups: Secretor [2]
Secretor Timepoint mean sd n
<dbl> <dbl> <dbl> <dbl> <int>
1 0 1 NA NA 2
2 0 6 NA NA 2
3 0 12 NA NA 2
4 0 18 NA NA 2
5 1 1 NA NA 2
6 1 6 NA NA 2
7 1 12 NA NA 2
8 1 18 NA NA 2
This is the warning I get:
In var(if (is.vector(x) || is.factor(x)) x else as.double(x), ... :
NAs introduced by coercion
Could anyone please provide advice as to why it works as a line of code, but not as a function?
Many thanks in advance.
@akrun's suggestion for how to immediately solve your question is right on.
An alternative is to use the nesting functionality of tidyr by returning a single element list which contains a data.frame of your results.
summary_function <- function(x) {
summary <- list(tibble(mean = mean(x, na.rm=T), sd = sd(x, na.rm=T), n = length(x[!is.na(x)])))
return(summary)
}
Then you can use across to do the same function to multiple columns:
dataSet %>%
group_by(Secretor, Timepoint) %>%
summarize(across(Gene1:Gene4, summary_function))
# A tibble: 8 x 6
# Groups: Secretor [2]
# Secretor Timepoint Gene1 Gene2 Gene3 Gene4
# <dbl> <dbl> <list> <list> <list> <list>
#1 0 1 <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]>
#2 0 6 <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]>
#3 0 12 <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]>
#4 0 18 <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]>
#5 1 1 <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]>
#6 1 6 <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]>
#7 1 12 <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]>
#8 1 18 <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]> <tibble [1 × 3]>
Now we can unnest those same columns using unnest with names_sep =:
dataSet %>%
group_by(Secretor, Timepoint) %>%
summarize(across(Gene1:Gene4, summary_function)) %>%
unnest(Gene1:Gene4, names_sep = "_")
# A tibble: 8 x 14
# Groups: Secretor [2]
# Secretor Timepoint Gene1_mean Gene1_sd Gene1_n Gene2_mean Gene2_sd Gene2_n Gene3_mean Gene3_sd Gene3_n
# <dbl> <dbl> <dbl> <dbl> <int> <dbl> <dbl> <int> <dbl> <dbl> <int>
#1 0 1 71.2 28.6 2 62.3 27.0 2 28.4 33.3 2
#2 0 6 5.40 7.43 2 58.6 29.1 2 37.0 33.9 2
#3 0 12 91.8 11.4 2 53.9 31.0 2 33.2 46.0 2
#4 0 18 51.5 65.0 2 65.3 40.2 2 63.8 32.7 2
#5 1 1 30.8 18.0 2 50.0 19.9 2 22.8 6.71 2
#6 1 6 63.9 49.2 2 59.9 41.8 2 30.9 39.5 2
#7 1 12 85.3 6.74 2 51.0 41.1 2 28.5 22.9 2
#8 1 18 41.7 44.8 2 80.2 24.0 2 64.7 17.4 2
## … with 3 more variables: Gene4_mean <dbl>, Gene4_sd <dbl>, Gene4_n <int>
This is a recent addition to tidyr and dplyr (version >=1.0.0), but can come handy.