calculating sum based on customed rules in the data frame - r

Question

Preferably using data.table in R: I want to calculate the sum of DIAM by ID, CYCLE # based on the following rules:
if any of the DIAM for particular subject cycle is presented as NE then the SUM cant be calculated (must return NA)
if any of the DIAM is presented as NA, then calculate the sum ignoring the NA (i.e. as if it is 0)
if none is NA, then calculate the sum as normal
Also i would like to substitute CYCLE number to numeric with BASELINE representing 0.
dfin <-
ID CYCLE NUM DIAM
1 BASELINE 1 8
1 BASLEINE 2 4
1 CYCLE 1 1 6
1 CYCLE 1 2 2
1 CYCLE 2 1 6
1 CYCLE 2 2 NE
1 CYCLE 3 1 6
1 CYCLE 3 2 NA
dfout <-
ID CYCLE SUM
1 0 12
1 1 8
1 2 NA
1 3 6
This need to be applied for every subject. There are many cycles there but this just an example.

Answer 1

Here is one option. Grouped by 'ID', and the matched index of 'CYCLE' (as showed in the expected output), change the "DIAM" values to NA if any of 'DIAM" have "NE", then summarise by taking the sum of 'DIAM' while making sure that if all of the values are NA return NA
library(tidyverse)
dfin %>%
group_by(ID, CYCLE = match(CYCLE, unique(CYCLE))-1) %>%
mutate(DIAM = as.numeric(replace(DIAM, any(DIAM== "NE"), NA))) %>%
summarise(SUM = NA^all(is.na(DIAM)) * sum(DIAM, na.rm = TRUE))
# A tibble: 4 x 3
# Groups: ID [?]
# ID CYCLE SUM
# <int> <dbl> <dbl>
#1 1 0 12
#2 1 1 8
#3 1 2 NA
#4 1 3 6
Or use an if/else condition after the group_by step
dfin %>%
group_by(ID, CYCLE = match(CYCLE, unique(CYCLE))-1) %>%
summarise(SUM = if("NE" %in% DIAM) NA else sum(as.numeric(DIAM), na.rm = TRUE))
Or using the same logic with data.table
library(data.table)
setDT(dfin)[, .(SUM = if("NE" %in% DIAM) NA_real_ else
sum(as.numeric(DIAM), na.rm = TRUE)), .(ID, CYCLE = rleid(CYCLE)-1)]
# ID CYCLE SUM
#1: 1 0 12
#2: 1 1 8
#3: 1 2 NA
#4: 1 3 6
data
dfin <- structure(list(ID = c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L),
CYCLE = c("BASELINE",
"BASELINE", "CYCLE 1", "CYCLE 1", "CYCLE 2", "CYCLE 2", "CYCLE 3",
"CYCLE 3"), NUM = c(1L, 2L, 1L, 2L, 1L, 2L, 1L, 2L), DIAM = c("8",
"4", "6", "2", "6", "NE", "6", NA)), row.names = c(NA, -8L),
class = "data.frame")

Answer 2

# Data created
dfin<-data.table("ID" = rep(x = 1,times = 8),"CYCLE" = c("BASELINE","BASELINE","CYCLE 1","CYCLE 1","CYCLE 2","CYCLE 2","CYCLE 3","CYCLE 3"),
"NUM" = rep(x = c(1,2),times = 4),"DIAM" = c(8,4,6,2,6,"NE",6,NA))
# CYCLE transformed
dfin[,CYCLE := as.numeric(ifelse(CYCLE == "BASELINE","0",
substr(x = CYCLE,start = 7,stop = 7)))]
# SUM computed
dfin2<-dfin[,.(SUM = if(CYCLE == 0){
NA_real_
} else if("NE" %in% DIAM){
NA_real_
} else {
sum(as.numeric(DIAM),na.rm = T)
}),by = c("ID","CYCLE")]
# IDs with CYCLE = 0 present have SUM updated to NA
dfin2[ID %in% ID[which(CYCLE == 0)],SUM := NA]
Hope this helps!