CODEWITHSUNDEEP
rfunctionloops
John_dydx
John_dydx

Reputation: 961

Looping through longitudinal dataset levels

I'm trying to select ids(patients) from each of 4 regions within the following 3 age groups: 2 - <4 yrs, 4 - <6 yrs, 6 - <8 yrs. The data is the longData given in the code below. I need to select only half of the ids that fit these age criteria and merge them in a separate data frame. I have written a function (subsetFunc) to do this individually for each region but I was hoping I could write a code using a for loop to apply a similar function to get this information from all four regions as a separate data frame (rather than applying the code 4 times for each region). I could simply apply the function 12 times and then rbind each output but this seems repetitive. In my real dataset, I have several regions and so it would be too tedious to apply this function say 60 times. Besides, this is the whole point of writing functions-to automate the repetitive things.

library(dplyr)
roundUp <- function(x) trunc(x+0.5)

set.seed(2016)

longData <- data.frame(patient=rep(paste(letters[1:20], sep = "_", 1:20), each=5), 
                   age=rep(runif(20, 1, 10), each = 5),var=runif(100, 50, 1000),
                   time=rep(1:5, 20), region = rep(c("EUROPE", "AFRICA", "ASIA",
                                                     "OCEANIA"), c(25, 25, 25, 25)))

subsetFunc <- function(df, group_in, seedNumb, a, b){

data <- filter(df, region == group_in, age >= a, age < b)
set.seed(seedNumb)
n <- roundUp(
data%>% filter(!duplicated(data["patient"])) %>% nrow()/2)
ids <- sample_n(unique(data[, c("patient", "region")]), n)$patient
fd <- data[data$patient %in% ids, ]
return(fd)

}

# this gives half of the children in Europe between ages 2 and < 4 yrs
subsetFunc(longData, "EUROPE", 1, 2, 4)

# Desired Output (tedious method for doing this)


# 2-<4
d_1  <- subsetFunc(longData, "EUROPE", 1, 2, 4)
d_2 <- subsetFunc(longData, "ASIA", 1, 2, 4)
d_3 <- subsetFunc(longData, "OCEANIA", 1, 2, 4)
d_4 <- subsetFunc(longData, "AFRICA", 1, 2, 4)

# 4- < 6
d_5 <- subsetFunc(longData, "EUROPE", 1, 4, 6)
d_6 <- subsetFunc(longData, "ASIA", 1, 4, 6)
d_7 <- subsetFunc(longData, "OCEANIA", 1, 4, 6)
d_8 <- subsetFunc(longData, "AFRICA", 1, 4, 6)

# 6- 8
d_9 <- subsetFunc(longData, "EUROPE", 1, 6, 8)
d_10 <- subsetFunc(longData, "ASIA", 1, 6, 8)
d_11 <- subsetFunc(longData, "OCEANIA", 1, 6, 8)
d_12 <- subsetFunc(longData, "AFRICA", 1, 6, 8)

finalLongData <- rbind(d_1, d_2, d_3, d_4, d_5, d_6, d_7, d_8, d_9,
                    d_10, d_11, d_12)

I started by writing something along the lines of:

for (i in levels(longData$cohort)){

dFrame <- subsetFunc(longData, i, 1, a = 0, b = 2)
return(dFrame)
}

I'm a little stuck at this stage, any assistance would therefore be highly appreciated.

Upvotes: 0

Views: 162

Answers (1)

bouncyball
bouncyball

Reputation: 10761

Alright, I came back to this post, and I think I might have cracked it. First, I needed to redefine your subsetFunc function:

subsetFunc <- function(df, seedNumb, a, b){
  data <- filter(df,  age >= a, age < b)
  set.seed(seedNumb)
  n <- ceiling(
    data%>% filter(!duplicated(data["patient"])) %>% nrow()/2)
  ids <- sample_n(unique(data[, c("patient", "region")]), n)$patient
  fd <- data[data$patient %in% ids, ]
  return(fd)
}

Then, I was able to put all the results together without having to use as many lines:

LongDataFinal <- 
do.call('rbind', 
           as.list(sapply(c(2,4,6), FUN = function(y) 
                   by(longData, longData$region, 
       FUN = function(x) subsetFunc(df = x, seedNumb = 1, a = y, b = y+2)))))

Upvotes: 2

Related Questions