Conglomorating hourly temperature data in R

Question

My goal is to find the minimum and maximum daily temperatures and add them to a data frame. My current data frame looks like the following:

ROW DATE_TIME  TEMP (DEG C)
1   5/1/1999   4.6
2   5/1/1999   3.8
3   5/1/1999   2.9

I am attempting to get the daily range of temperature using this data, but the main issue I run into is having "non-standard" dates. The dataset I'm using is several thousand data points long, so I would like to have a code that does max-min for every 24 rows in order to get the daily variation in temperature.

Thank you!

Each 24 rows is day? could I ignore the DATE_TIME variable? – DJV May 29 '18 at 16:06 — DJV, May 29 '18 at 16:06
Yes. I think that would work. @DJV – Will Kirkpatrick May 29 '18 at 16:11 — Will Kirkpatrick, May 29 '18 at 16:11

DJV · Accepted Answer · 2018-05-31T12:05:03.797

If you want to calculate it using a running window you can use the function gtools::running() and set the by() and width() arguments to 24.

require(tidyverse)
require(gtools)

set.seed(123)
df <- data.frame(row = c(seq(1, 24*5, by = 1)), 
                 date = as.Date(c(
                   rep(c("02/25/92"), 24), 
                   rep(c("02/26/92"), 24),
                   rep(c("02/27/92"), 24),
                   rep(c("02/28/92"), 24), 
                   rep(c("02/29/92"), 24)), 
                   format = "%m/%d/%y"),
                 temp = rnorm(24*5, mean = 5, sd = 5)) 

#Function to calculate the min. and max. of a vector/column 
MinMaxFunction <- function(x) {
  return(data.frame(min = min(x, na.rm = TRUE), 
             max = max(x, na.rm = TRUE)))
}

#Calculating the running min. max. 
dfRunningMean <- running(df$temp,
                         fun = MinMaxFunction,
                         by = 24, 
                         width = 24) %>%
  t() %>% 
  as.data.frame()

dfRunningMean

             min      max
1:24   -4.833086 13.93457
25:48  -3.433467 15.84478
49:72  -6.545844 15.25042
73:96  -1.103589 11.80326
97:120  -3.33971 15.93666

Or, you can do it with the tidyverse approach, and calculate the min./max. for each date.

require(tidyverse)


df %>% 
  group_by(date) %>% 
  summarise(min = min(temp, na.rm = TRUE), 
            max = max(temp, na.rm = TRUE))

  date         min   max
  <date>     <dbl> <dbl>
1 1992-02-25 -4.83  13.9
2 1992-02-26 -3.43  15.8
3 1992-02-27 -6.55  15.3
4 1992-02-28 -1.10  11.8
5 1992-02-29 -3.34  15.9

The issue that I am seeing is that I do not need the mean of every day value, but the max-min of every 24. It is essentially like every 24 rows is a subset, and I need the maximum-minimum of each. Maybe this cannot be done? — Will Kirkpatrick, May 31 '18 at 11:43
Glad I could help and if this answers your question, please check "accept" answer :) — DJV, May 31 '18 at 13:18

Conglomorating hourly temperature data in R

1 Answers1