用R中2个最接近的行的平均值替换缺失值

Question

I have a datatable with missing value, and I want to replace it with the average of 2 nearest row.

library(data.table)
A <- data.table(id = c(1:10),
                Value = c(1:3,NA,5:10))

> A
    id Value
 1:  1     1
 2:  2     2
 3:  3     3
 4:  4    NA
 5:  5     5
 6:  6     6
 7:  7     7
 8:  8     8
 9:  9     9
10: 10    10

For example, I want the NA to be replaced by the mean of row3 and row5, which is 4.

Answer 1

na.approx in the zoo package does that. If there can be leading or trailing NA values and you want to:

• extend the nearest non-NA values add the rule = 2 argument to na.approx or
• leave those as NA add the na.rm = FALSE argument to na.approx .

See ?na.approx for further arguments. Other possibilities from the same package include na.spline (fill in with cubic spline fit), na.aggregate (mean of all non-NA values), na.locf (last value carried forward) and na.StructTS (seasonal Kalman filter).

library(zoo)

A[, list(Value = na.approx(Value))]

giving:

    Value
 1:     1
 2:     2
 3:     3
 4:     4
 5:     5
 6:     6
 7:     7
 8:     8
 9:     9
10:    10

Answer 2

I have made a function that work with more than one continuous NA in your data table.

library(data.table)
A <- data.table(id = c(1:11),
            Value = c(1,5:6,NA,10:12,NA,NA,NA,6))



library(dplyr)
# Finding the maximum length of a stretch of contiguous NA's in the column
a<- max(diff(which(!is.na(A$Value)))-1)

# Repeating the for loop "a" times and breaking when all NA's have been filled
repeat{
for(i in 1:a){
A$Value[which(A$Value%in%NA)] <- ((lag(A$Value, 1)+lead(A$Value, i))/2)[which(A$Value%in%NA)]
}
if(any(is.na(A$Value)) ==FALSE) { break }
}

For every NA in the vector, the function inside the for loop calculates the mean of the value previous to NA and the next available one.

This is definitely not the most elegant or efficient solution as there is a lot of repetition but i believe it works with more than one NA in the manner you desire.