R语言时间序列线性插值求助：填充体温数据缺失值

Hey there! Let's work through this problem step by step since you're new to R—no worries, we'll get your temperature data interpolated correctly. First, let's break down what's going wrong with your current code, then dive into two straightforward solutions that fit your needs (handling single NAs and skipping continuous NAs longer than 10 minutes).

What's Wrong With Your Custom Code? #

Your ifelse approach has a key syntax issue: in R, writing df$var1[+1] doesn't mean "the next row"—it actually just refers to the first row of the column! Similarly, df$var1[-1] removes the first row entirely (shifting the whole column up by one), which isn't what you want for per-row NA replacement. Since ifelse works vectorially (on entire columns at once), this approach can't target the specific previous/next values around each NA, especially not for continuous missing stretches.

Solution 1: Use zoo Package (Simplest for Your Needs) #

The zoo package has a built-in function na.approx() that was made exactly for this kind of time-series interpolation. It handles linear interpolation automatically, and lets you set a limit for how many consecutive NAs you want to interpolate (perfect for skipping stretches longer than 10 minutes).

Step 1: Install and Load the Package

install.packages("zoo")  # Run once to install
library(zoo)             # Load every time you use it

Step 2: Interpolate Your Data

Assuming your temperature column is named temp in your data frame df, run this:

# Interpolate NAs, but leave stretches longer than 10 consecutive NAs as NA
df$temp_interpolated <- na.approx(df$temp, maxgap = 10)

• maxgap = 10: This tells R to only interpolate if there are 10 or fewer consecutive NAs. Any longer stretches stay as NA, which matches your requirement.
• The function uses linear interpolation by default, so single NAs will be filled with the midpoint of the nearest non-NA values before/after.

Solution 2: Base R (No Extra Packages) #

If you prefer to stick with base R, you can use the approx() function combined with rle() to handle the long consecutive NA rule.

Step 1: Create a Time Index

Since your data is collected every minute over 8 hours, we can use row numbers as a time index (1 to 480 minutes):

df$minute <- 1:nrow(df)

Step 2: Run Linear Interpolation

# Get interpolated values for all minutes
interpolated <- approx(
  x = df$minute[!is.na(df$temp)],  # Non-NA time points
  y = df$temp[!is.na(df$temp)],    # Corresponding temperature values
  xout = df$minute,                # All time points to fill
  method = "linear"                # Use linear interpolation
)$y

# Add interpolated values to your data frame
df$temp_interpolated <- interpolated

Step 3: Keep Long Consecutive NAs as NA

Now we'll identify stretches of 11+ consecutive NAs and set those back to NA:

# Calculate runs of NAs in the original temp column
na_runs <- rle(is.na(df$temp))

# Create a logical vector marking long NA stretches (>10)
long_na_mask <- inverse.rle(
  list(
    lengths = na_runs$lengths,
    values = na_runs$values & na_runs$lengths > 10
  )
)

# Set long NA stretches back to NA in the interpolated column
df$temp_interpolated[long_na_mask] <- NA

Example to Test It Out #

Let's make a sample dataset to see how this works:

# Simulate 8 hours of temperature data with random NAs
set.seed(123)  # For reproducible results
df <- data.frame(
  minute = 1:480,
  temp = rnorm(480, mean = 36.5, sd = 0.2)
)

# Insert 20 single NAs and one 15-minute consecutive NA stretch
df$temp[sample(1:480, 20)] <- NA
df$temp[100:114] <- NA

If you run the zoo solution on this, you'll see:

• All single NAs are filled with linear interpolations
• The 15-minute NA stretch (100-114) stays as NA (since 15 > 10)

Final Notes #

As a new R user, the zoo package method is definitely the way to go—it's concise, readable, and handles all the edge cases (like long NA stretches) without extra code. Avoid writing custom ifelse logic for missing values unless absolutely necessary; R has tons of built-in tools for data cleaning that save time and reduce errors.

内容的提问来源于stack exchange，提问作者Eve