CODEWITHSUNDEEP
rdplyrdata.tabledtplyr
EML
EML

Reputation: 671

Create tables by using data.table and a for loop for multiple columns

I need to speed up code using data.table. I am getting stuck on how to reference variables that are being indexed from a vector.

data:

df <- data.frame(
  id=c(1,1,1,2,2,2,3,3,3),
  year=as.character(c(2014, 2015, 2016, 2015, 2015, 2016, NA, NA, 2016)),
  code=c(1,2,2, 1,2,3, 3,4,5),
  dv1=1:9,
  dv2=2:10
) %>% as.data.table()

dtplyr code:

cols <- c("dv1", "dv2")

test <- function(data, columns, group) {
for(i in seq_along(columns)) {
 sub1 <- df %>% 
   select("id", columns[i], group) %>%
   group_by(.data[[group]]) %>%
   summarise(mean=mean(.data[[columns[i]]], na.rm=T), sd=sd(.data[[columns[i]]], na.rm=T)) %>%
   ungroup() %>%
   as_tibble() 
 print(sub1)
}
}

data.table attempt:

test <- function(data, columns, group) {
  for(i in seq_along(columns)) {
    sub1 <- df %>% 
      .[, .(id, columns[i], group)] %>%
      .[, .(mean(.data[[columns[i]]], na.rm=T), sd=sd(.data[[columns[i]]], na.rm=T)), by=.data[[group]]] %>%
      as_tibble() 
    print(sub1)
  }
}

test(data=df, columns=cols, group="year")

This works on a single variable:

df %>% 
  .[, .(id, dv1, year)] %>%
  .[, .(mean(dv1, na.rm=T), sd=sd(dv1, na.rm=T)), by=year] %>%
  as_tibble()

Upvotes: 0

Views: 948

Answers (2)

Ronak Shah
Ronak Shah

Reputation: 388947

  • .data is not used in data.table
  • You don't need select here and that is why you also don't need .[, .(id, columns[i], group)] in data.table version.
  • You can use get to get column values based on string.

Since this is just an example I have not tried to simplify the loop so that you can add more complicated stuff in there later.

library(data.table)

cols <- c("dv1", "dv2")

test <- function(data, columns, group) {
  for(i in columns) {
    sub1 <-df[, .(mean(get(i), na.rm=T), sd=sd(get(i), na.rm=T)), by=year]
    print(sub1)
  }
}

test(data=df, columns=cols, group="year")

#   year   V1    sd
#1: 2014 1.00    NA
#2: 2015 3.67 1.528
#3: 2016 6.00 3.000
#4: <NA> 7.50 0.707

#   year   V1    sd
#1: 2014 2.00    NA
#2: 2015 4.67 1.528
#3: 2016 7.00 3.000
#4: <NA> 8.50 0.707

Upvotes: 2

Sirius
Sirius

Reputation: 5429

This likely will require a fairly unintuitive as.list/unlist construction:


df <- data.frame(
  id=c(1,1,1,2,2,2,3,3,3),
  year=as.character(c(2014, 2015, 2016, 2015, 2015, 2016, NA, NA, 2016)),
  code=c(1,2,2, 1,2,3, 3,4,5),
  dv1=1:9,
  dv2=2:10
) %>% as.data.table()

summary.func <- function(x) {
    list( mean=mean(x), sd=sd(x) )
}

df[, as.list(unlist(lapply(.SD, summary.func))), by=group, .SDcols=cols ]

It produces:


   year dv1.mean    dv1.sd dv2.mean    dv2.sd
1: 2014 1.000000        NA 2.000000        NA
2: 2015 3.666667 1.5275252 4.666667 1.5275252
3: 2016 6.000000 3.0000000 7.000000 3.0000000
4: <NA> 7.500000 0.7071068 8.500000 0.7071068

Upvotes: 1

Related Questions