Creating a time series plot and converting numeric data to date - r

Question

I want to create a plot of time per temperature in 2 sites. I have data of the temperature each 10 minutes a day from february to april and I need daily cycles of hourly averages of temperature to plot.
I calculated the mean temperature for hour a day and try to create a plot with geom_plot and geopm_line of different ways.
data <- read.xlsx("temperatura.xlsx", 1)
data <- data %>% mutate (month = as.factor(month), month = as.factor (month), day = as.factor(day), h = as.factor(h), min = as.factor(min))
head (data)
month day h min t.site1 t.site2
2 1 0 0 15.485 16.773
2 1 0 10 15.509 16.773
2 1 0 20 15.557 16.773
2 1 0 30 15.557 16.773
2 1 0 40 15.605 16.773
2 1 0 50 15.605 16.773
str(data)
'data.frame': 12816 obs. of 6 variables:
$ month : Factor w/ 3 levels "2","3","4": 1 1 1 1 1 1 1 1 1 1 ...
$ day : Factor w/ 31 levels "1","2","3","4",..: 1 1 1 1 1 1 1 1 1 1 ...
$ h : Factor w/ 24 levels "0","1","2","3",..: 1 1 1 1 1 1 2 2 2 2 ...
$ min : Factor w/ 6 levels "0","10","20",..: 1 2 3 4 5 6 1 2 3 4 ...
$ t.site1: num 15.5 15.5 15.6 15.6 15.6 ...
$ t.site2: num 16.8 16.8 16.8 16.8 16.8 ...
hour <- group_by(data, month, day, h)
mean.h.site1 <- summarize(hour, mean.h.site1 = mean(t.site1))
t1 <- ggplot (data = mean.h.site1, aes(x=h, y=mean.h.site1)) +
geom_line()
t2 <- ggplot(data = mean.h.site1, aes(x=h, y=mean.h.site1, group = month))+
geom_line() +
geom_point()
t3 <- ggplot (data = mean.h.site1, aes(x=day, y=mean.h.site1, group=1))+
geom_point()
I expect the output of the variability of temperature across the time for each site, but the actual output show temperature variability during each day.

Answer 1

It's interesting that your data is showing month, day and hour as factor. Is it possible that there are some character values somewhere in that column when you read the data? It's very unusual to see numbers stored as factor in that fashion.
I'll do 4 things:
Convert factors to numbers
Convert numbers to dates
Convert a wide table to a long one, and finally
plot the temps against a real date
# Load packages and data
library(data.table) # for overall fast data processing
library(lubridate) # for dates wrangling
library(ggplot2) # plotting
dt <- fread("month day h min t.site1 t.site2
2 1 0 0 15.485 16.773
2 1 0 10 15.509 16.773
2 1 0 20 15.557 16.773
2 1 0 30 15.557 16.773
2 1 0 40 15.605 16.773
2 1 0 50 15.605 16.773")
# Convert factors to numbers (I actuall didn't run this because I just created the data.table, but it seems you'll need to do it):
dt[, names(dt)[1:4] := lapply(.SD, function(x) as.numeric(as.character(x)), .SDcols = 1:4]
# Create proper dates. We'll consider all dates occurring in 2019.
dt[, date := ymd_hm(paste0("2019/", month, "/", day, " ", h, ":", min))]
# convert wide data to long one
dt2 <- melt(dt[, .(date, t.site1, t.site2)], id.vars = "date")
# plot the data
ggplot(dt2, aes(x = date, y = value, color = variable))+geom_point()+geom_path()

Answer 2

You could paste the time columns together and convert them as.POSIXct.
As #PavoDive already pointed out we'll need numeric time columns. Check your code that produced the data or transform to numeric with d[1:4] <- Map(function(x) as.numeric(as.character(x)), d[1:4]).
Now paste the rows with apply, convert as.POSIXct, and cbind it to the remainder. The sprintf looks first that all values have the same digits before pasting.
d2 <- cbind(time=as.POSIXct(apply(sapply(d[1:4], sprintf, fmt="%02d"), 1, paste, collapse=""),
format="%m%d%H%M"),
d[5:6])
Plots nicely, here in base R:
with(d2, plot(time, t.site1, ylim=c(15, 17), xaxt="n",
xlab="time", ylab="value", type="b", col="red",
main="Time series"))
with(d2, lines(time, t.site2, type="b", col="green"))
mtext(strftime(d2$time, "%H:%M"), 1, 1, at=d2$time) # strftime gives the desired formatting
legend("bottomright", names(d2)[2:3], col=c("red", "green"), lty=rep(1, 2))
Data
d <- structure(list(month = structure(c(1L, 1L, 1L, 1L, 1L, 1L), .Label = "2", class = "factor"),
day = structure(c(1L, 1L, 1L, 1L, 1L, 1L), .Label = "1", class = "factor"),
h = structure(c(1L, 1L, 1L, 1L, 1L, 1L), .Label = "0", class = "factor"),
min = structure(1:6, .Label = c("0", "10", "20", "30", "40",
"50"), class = "factor"), t.site1 = c(15.485, 15.509, 15.557,
15.557, 15.605, 15.605), t.site2 = c(16.773, 16.773, 16.773,
16.773, 16.773, 16.773)), row.names = c(NA, -6L), class = "data.frame")

Answer 3

I'm assuming that you needed the actual output showing temperature variability by hour for each day in the same plot?
EDITED:
I have updated the code to generate a day worth of data. And, also generate the chart.
library(tidyverse)
library(lubridate)
df <- data_frame(month = rep(2, 144),
day = rep(1, 144),
h = rep(0:24, each = 6, len = 144),
min = rep((0:5)*10,24),
t.site1 = rnorm(n = 144, mean = 15.501, sd = 0.552),
t.site2 = rnorm(n = 144, mean = 16.501, sd = 0.532))
df %>%
group_by(month, day, h) %>%
summarise(mean_t_site1 = mean(t.site1), mean_t_site2 = mean(t.site2)) %>%
mutate(date = ymd_h(paste0("2019-",month,"-",day," ",h))) %>%
ungroup() %>%
select(mean_t_site1:date) %>%
gather(key = "site", value = "mean_temperature", -date) %>%
ggplot(aes(x = date, y = mean_temperature, colour = site)) +
geom_line()
Could you verify if this is the output you need?