Calculate a new median for each row by id

I know this may be very simple but I can't solve it.

I have the following df:

Input Data

df<-data.frame(id=c(1,2,3,3,3,4, 4, 4, 4, 4, 4), value = c(956, 986, 995, 995, 986, 700, 600, 995, 956, 1000, 986))

Desired Result

df<-data.frame(id=c("1","2","3","3","3","4", "4", "4", "4", "4", "4"), value = c("956", "986", "995", "995", "986", "700", "600", "995", "956", "1000", "986"), median = c("956", "986","995","995", "995", "700","650","700","828", "956", "971"))

This is, to calculate the median value for each row (taking into account the different ID's). In each row a new value will be added and then a new median should be calculated

Output Data

library(dplyr)
w = df %>%
group_by(id) %>%
mutate(median = median(value, na.rm =TRUE)) %>%
select (median)
df$median <- w[,2]


df<-data.frame(id=c("1","2","3","3","3","4", "4", "4", "4", "4", "4"), value = c("956", "986", "995", "995", "986", "700", "600", "995", "956", "1000", "986"), median = c("956", "986","995","995", "995", "971","971","971","971", "971", "971"))

Solution

You can use zoo::rollapplyr to calculate a rolling median:

library(tidyverse);
library(zoo);
df %>%
    group_by(id) %>%
    mutate(
        median = rollapplyr(value, seq_along(value), median))
## A tibble: 11 x 3
## Groups:   id [4]
#      id value median
#   <dbl> <dbl>  <dbl>
# 1    1.  956.   956.
# 2    2.  986.   986.
# 3    3.  995.   995.
# 4    3.  995.   995.
# 5    3.  986.   995.
# 6    4.  700.   700.
# 7    4.  600.   650.
# 8    4.  995.   700.
# 9    4.  956.   828.
#10    4. 1000.   956.
#11    4.  986.   971.

Sample data

df <- data.frame(
    id = c(1,2,3,3,3,4, 4, 4, 4, 4, 4), 
    value = c(956, 986, 995, 995, 986, 700, 600, 995, 956, 1000, 986))