Add multiple columns to data.frame using a function that returns multiple outputs in dplyr

Question

I am calculating median and values for different quantiles for a continuous variable. I want to add all the columns in a single step. Is this possible to do this. Following is a reproducible example.

df <- data.frame(group = rep(c('group1','group2'),50),
             x = rnorm(100), 
             y = rnorm(100))
df %>% 
gather('variable','value', -group) %>% 
group_by(group, variable) %>% 
summarise(median = round(quantile(value,0.5, na.rm = T),2),
          iqr25 = round(quantile(value,0.25, na.rm = T),2),
          iqr75 = round(quantile(value,0.75, na.rm = T),2))

OUTPUT

# A tibble: 4 x 5
# Groups:   group [2]
  group  variable median iqr25 iqr75
  <fct>  <chr>     <dbl> <dbl> <dbl>
1 group1 x          0.06 -0.74  1.04
2 group1 y         -0.36 -1.03  0.45
3 group2 x         -0.04 -0.85  0.62
4 group2 y          0.06 -0.56  0.89

Can this summarise step be done without writing the quantile function 3 times.

I did a work around using this. But is there a nice way to do this.

df %>% 
gather('variable','value', -group) %>% 
group_by(group, variable) %>% 
summarise(s = toString(round(quantile(value, c(0.25,0.5,0.75),na.rm = T),2))) %>% 
separate(s, into = c('q25','median','q75'), sep = ',')

Richard Telford · Accepted Answer · 2019-09-02 08:08:12Z

1

You can nest the data after the group_by and then map to quantile

df %>% 
  gather('variable','value', -group) %>% 
  group_by(group, variable) %>% 
  nest() %>% 
  mutate(quant = map(data, ~quantile(.$value, probs = c(0.25, 0.5, 0.75))),
         quant = map(quant, t),
         quant = map(quant, as.data.frame),
         quant = map(quant, setNames, c("iqr25", "median", "iqr75")),

         ) %>% 
  unnest(quant) %>% 
  select(-data)

# A tibble: 4 x 5
  group  variable  iqr25  median iqr75
  <fct>  <chr>     <dbl>   <dbl> <dbl>
1 group1 x        -0.876 -0.173  0.471
2 group2 x        -0.372  0.0507 0.519
3 group1 y        -0.785 -0.109  0.618
4 group2 y        -0.944 -0.117  0.647

edited Sep 2, 2019 at 8:08

answered Sep 2, 2019 at 8:03

Richard Telford

9,9836 gold badges43 silver badges53 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Koundy Over a year ago

Nice! Can we add column names to this?

Richard Telford Over a year ago

see edit for names - could also use dplyr::rename after the unnest

Koundy Over a year ago

Thanks this nest followed by map approach helps in a lot of other things as well. Will try to incorporate these in routine workflow

teofil · Accepted Answer · 2019-09-02 08:25:54Z

0

Another approach using nest:

df %>%
  gather('variable', 'value', -group) %>%
  group_by(group, variable) %>%
  nest() %>%
  mutate(quants = map(data, function(x) 
    quantile(x$value, c(0.25,0.5,0.75)))) %>%
  unnest(quants) %>%
  group_by(group, variable) %>%
  mutate(case = c("iqr25", "median" , "iqr75")) %>%
  spread(case, quants) %>% 
  mutate_if(is.numeric, round, 2)

# A tibble: 4 x 5
# Groups:   group, variable [4]
  group  variable iqr25 iqr75 median
  <fct>  <chr>    <dbl> <dbl>  <dbl>
1 group1 x        -0.5   0.7    0.09
2 group1 y        -0.54  0.7    0.1 
3 group2 x        -0.59  0.61  -0.06
4 group2 y        -0.89  0.35  -0.11

edited Sep 2, 2019 at 8:25

answered Sep 2, 2019 at 8:18

teofil

2,3941 gold badge10 silver badges17 bronze badges

4 Comments

Koundy Over a year ago

This code actually looks more verbose and messy than using quantile function 3 times in summarize !

teofil Over a year ago

Well, I for one think that your original code is cleanest and most readable, even though you are calling quantile 3 times. The alternatives are performing unnecessary transformations on the whole dataset to achieve the same thing in more lines of code. I agree that Richard's solution is cleaner than mine, but still we are replacing 3 calls to quantile with 4 calls to map?!

Koundy Over a year ago

yes. that is the reason I didn't accept his answer.

Richard Telford Over a year ago

You could just use one call to map, but the contents then get a bit messy -

map(data, ~{quantile(.$value, probs = c(0.25, 0.5, 0.75))) %>% t() %>% as.data.frame() %>% setNames, c("iqr25", "median", "iqr75"))})

Collectives™ on Stack Overflow

Add multiple columns to data.frame using a function that returns multiple outputs in dplyr

2 Answers 2

3 Comments

4 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

4 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related