基于R语言计算24小时周期内TRUE事件的发生频率

Hey there! To compute the frequency of AbovePeak = TRUE events within each 24-hour period using your dataset, here's a step-by-step approach using R's dplyr and lubridate packages:

1. Load Required Libraries #

First, make sure you have these packages installed (if not, run install.packages(c("dplyr", "lubridate"))), then load them:

library(dplyr)
library(lubridate)

2. Prepare Your Data #

We'll start by extracting the date component from your DAT datetime column to define our 24-hour periods. By default, this uses calendar days (midnight to midnight), but I'll show how to adjust for custom windows later.

Assuming your dataset is stored in a data frame called df:

df <- df %>%
  mutate(calendar_date = date(DAT)) # Pulls the YYYY-MM-DD date from the datetime stamp

3. Calculate Frequency Metrics #

Now group the data by year, station, and calendar date to compute key frequency stats:

• Total observations per 24-hour period
• Number of AbovePeak = TRUE events
• Relative frequency (proportion of observations that are peak events)

daily_peak_frequency <- df %>%
  group_by(YEAR, STATION, calendar_date) %>%
  summarise(
    total_observations = n(),
    peak_event_count = sum(AbovePeak, na.rm = TRUE),
    peak_frequency = peak_event_count / total_observations
  ) %>%
  ungroup()

Quick Notes on the Code: #

• sum(AbovePeak) works because R treats TRUE as 1 and FALSE as 0—perfect for counting events.
• na.rm = TRUE ensures any missing values in AbovePeak don't mess up the calculation.
• Use peak_event_count if you just want the raw number of peak events per day. Use peak_frequency if you want the ratio of peak events to total observations (great for irregular observation intervals).

4. Custom 24-Hour Windows (If Needed) #

If your 24-hour period doesn't line up with calendar days (e.g., runs from 7 AM to 7 AM the next day), shift the datetime to group correctly:

df <- df %>%
  mutate(
    # Shift datetime back by 7 hours so 7 AM becomes midnight of our "window date"
    shifted_datetime = DAT - hours(7),
    window_date = date(shifted_datetime)
  )

# Now group by window_date instead of calendar_date
custom_window_frequency <- df %>%
  group_by(YEAR, STATION, window_date) %>%
  summarise(
    total_observations = n(),
    peak_event_count = sum(AbovePeak, na.rm = TRUE),
    peak_frequency = peak_event_count / total_observations
  ) %>%
  ungroup()

Example Output #

Your final data frame will look something like this (truncated):

This tells you that on January 1, 1993, at station COOP:047821, there was 1 peak event out of 8 total observations—so a 12.5% frequency of peak events that day.

内容的提问来源于stack exchange，提问作者Michaela