Apply custom function to specific row/column - r

Question

I'm trying to solve a much larger problem using this basic example. I need to apply a function based on the location from which() because I need to know the year from df1 where the value is NA or >= 150. Then I subset df2, get the mean, and return it to the exact row. Right now I'm using a for() loop and need something much faster as the data I have is very large. Is there a common way to do this?
dput:
df1 <- structure(list(id = c("USC00031632", "USC00031632", "USC00031632",
"USC00031632", "USC00031632", "USC00031632", "USC00031632", "USC00031632",
"USC00031632", "USC00031632"), element = c("TMAX", "TMIN", "TMAX",
"TMIN", "TMAX", "TMIN", "TMAX", "TMIN", "TMAX", "TMIN"), year = 1900:1909,
month = c(1, 1, 2, 2, 3, 3, 4, 4, 5, 5), day = c(1, 1, 1,
1, 1, 1, 1, 1, 1, 1), value = c(30.02, NA, 37.94, 10.94,
NA, 28.04, 64.94, 41, 200, 51.08)), .Names = c("id", "element",
"year", "month", "day", "value"), row.names = c(NA, -10L), class = c("tbl_df",
"data.frame"))
df2 <-structure(list(id = c("USC00031632", "USC00031632", "USC00031632",
"USC00031632", "USC00031632", "USC00031632", "USC00031632", "USC00031632",
"USC00031632", "USC00031632"), element = c("TMAX", "TMIN", "TMAX",
"TMIN", "TMAX", "TMIN", "TMAX", "TMIN", "TMAX", "TMIN"), year = 1900:1909,
month = c(1, 1, 2, 2, 3, 3, 4, 4, 5, 5), day = c(1, 1, 1,
1, 1, 1, 1, 1, 1, 1), value = c(30.02, 10.94, 37.94, 10.94,
12, 28.04, 64.94, 41, 82.04, 51.08)), row.names = c(NA, -10L
), class = c("tbl_df", "data.frame"), .Names = c("id", "element",
"year", "month", "day", "value"))
Code:
library(dplyr)
check <- function(df, yr){
df_d <- filter(df, year == yr)
m <- mean(df_d$value)
return(m)
}
for (i in which(is.na(df1$value) | df1$value >= 150)){
df1[i,6] <- check(df = df2, yr = as.numeric(df1[i,3]) )
}

Answer 1

I would recommend the efficient binary join from data.table combined with modification in place (using the :=) while specifying by = .EACHI (in order to calculate the mean for each group separately).
library(data.table)
setDT(df1)[setDT(df2),
value := ifelse(is.na(value) | value >= 150, mean(i.value), value),
on = "year",
by = .EACHI]
df1
# id element year month day value
# 1: USC00031632 TMAX 1900 1 1 30.02
# 2: USC00031632 TMIN 1901 1 1 10.94
# 3: USC00031632 TMAX 1902 2 1 37.94
# 4: USC00031632 TMIN 1903 2 1 10.94
# 5: USC00031632 TMAX 1904 3 1 12.00
# 6: USC00031632 TMIN 1905 3 1 28.04
# 7: USC00031632 TMAX 1906 4 1 64.94
# 8: USC00031632 TMIN 1907 4 1 41.00
# 9: USC00031632 TMAX 1908 5 1 82.04
# 10: USC00031632 TMIN 1909 5 1 51.08
Alternatively, we could do this in two steps in order to try avoiding the ifelse overhead in each step
setDT(df1)[setDT(df2), value2 := i.value, on = "year"]
df1[is.na(value) | value >= 150, value := mean(value2), by = year]
df1
# id element year month day value value2
# 1: USC00031632 TMAX 1900 1 1 30.02 30.02
# 2: USC00031632 TMIN 1901 1 1 10.94 10.94
# 3: USC00031632 TMAX 1902 2 1 37.94 37.94
# 4: USC00031632 TMIN 1903 2 1 10.94 10.94
# 5: USC00031632 TMAX 1904 3 1 12.00 12.00
# 6: USC00031632 TMIN 1905 3 1 28.04 28.04
# 7: USC00031632 TMAX 1906 4 1 64.94 64.94
# 8: USC00031632 TMIN 1907 4 1 41.00 41.00
# 9: USC00031632 TMAX 1908 5 1 82.04 82.04
# 10: USC00031632 TMIN 1909 5 1 51.08 51.08
You can get rid of value2 afterwards if you wish using df1[, value2 := NULL]