R语言小数秒舍入问题：10Hz时间序列生成异常求助

Hey there! I totally get the frustration when your time series doesn’t line up as expected—those tiny floating-point gremlins can be tricky. Let’s break down why this is happening and how to fix it.

The Root Cause #

Your issue boils down to binary floating-point precision: 0.1 can’t be represented exactly in binary (it’s an infinite repeating fraction, like 1/3 in decimal). When you add 0.1 repeatedly to build your time sequence, those tiny errors accumulate, leading to unexpected values at positions 2, 4, etc. Other decimals might work because they can be represented exactly in binary, but 0.1 isn’t one of them.

Practical Fixes #

1. Use Integer Microsecond Increments (No External Packages)

POSIXct timestamps in R are stored as the number of seconds (with microsecond precision) since the Unix epoch. Instead of adding 0.1 seconds directly, calculate the increment in microseconds (0.1s = 100,000 microseconds) and add those integers—this avoids floating-point drift entirely.

Example code:

# Define your start time
start_time <- as.POSIXct("2024-01-01 00:00:00", tz = "UTC")

# Number of samples you want (adjust as needed)
n_samples <- 5

# Generate the fixed sequence using integer microsecond increments
fixed_times <- start_time + (0:(n_samples - 1)) * 100000

# Verify the exact values (use digits=15 to see full precision)
print(fixed_times, digits = 15)

2. Use the lubridate Package (More Intuitive)

If you prefer a cleaner, human-readable approach, the lubridate package handles time increments with built-in functions that avoid precision issues. Even in R 3.4.1, this package should work smoothly.

Example code:

# Install if you haven't already (run once)
# install.packages("lubridate")

library(lubridate)

start_time <- ymd_hms("2024-01-01 00:00:00", tz = "UTC")
n_samples <- 5

# Generate sequence with 100ms increments (0.1s)
fixed_times <- start_time + milliseconds(0:(n_samples - 1) * 100)

# Check the results
print(fixed_times, digits = 15)

3. Validate Your Original Sequence

To confirm the floating-point error is the culprit, run this on your original sequence:

# Your original sequence code might look like this
original_times <- seq(start_time, by = 0.1, length.out = n_samples)

# Print with full precision to see the hidden error
print(original_times, digits = 15)

You’ll notice values like 2024-01-01 00:00:00.10000000149 instead of the exact 0.1 increment—this is where your unexpected values come from.

Bonus Tip #

If possible, consider upgrading your R version (3.4.1 is quite old at this point). Newer R releases have improved handling of floating-point operations and time series, which can prevent these issues from popping up in the first place. That said, the fixes above will work perfectly in your current setup.

内容的提问来源于stack exchange，提问作者Dr Simon Holgate