group by - roll-up record, max of each column, group_by R -

this seems simple, , have solution, it's kinda time consuming since have lot of columns. have looked @ other solutions, it's been different (aggregate 1 column, mutate columns etc). in sql select pat_id, max(x), max(y), max(z) table_name group pat_id.

i have data set looks (but more columns):

dt <- data.frame(   pat_id = c('p','p','p','a','a','a'),   x = c(1,na,na, 1,na,na),   y = c(na,2,na,na,1,na),   z = c(na,na,1,na,na,0) ) 

so summarize , combine results:

results_x  <-dt %>%                                 group_by(pat_id ) %>%                           summarise(x = max(x, na.rm=true))   results_y  <-dt %>%                                group_by(pat_id ) %>%                          summarise(y = max(y, na.rm=true))   results_z  <-dt %>%                                group_by(pat_id ) %>%                          summarise(z = max(z, na.rm=true))   resulted <- left_join(results_x, results_y ) resulted <- left_join(resulted, results_z) 

my output "roll-up" record max value each column per pat_id:

   myresult <- data.frame(     pat_id = c('p','a'),     x = c(1,1),     y = c(2,1),     z = c(1,0)     ) 

i'm sure there's better way this, how?

this can done summarize_all in dplyr. here go

library(dplyr) dt %>% group_by(pat_id) %>% summarize_all(max, na.rm=t) #   pat_id     x     y     z #   <fctr> <dbl> <dbl> <dbl> # 1          1     1     0 # 2      p     1     2     1