R env (#8)

* Create R environment using 'renv' package. Created a basic snapshot using

dplyr, magrittr and janitor

Started the process of replication the Basel weather python notebook.

* Updated the basel_weather.csv. Added the following analyses:

	Clean columns
	Create various date variables
	Plot the hourly and daily weather
	Added spider plot for cyclical seasonality
	Create moving averages 7, 14 and 365 days
	Added kernel smoother with bandwidth of 30, 90

* Bandwidth = 2*sd for the ksmooth function. Adjusted the R script accordingly.

* Updated the working directory path.

Update the rlock file to inclue the "here" package.

* Changed renv path to the root directory.

* Finalised folder structure.

Co-authored-by: Aaron Pickering <[email protected]>

.Rprofile

@@ -0,0 +1 @@
+source("renv/activate.R")

.gitignore

@@ -133,4 +133,5 @@ dmypy.json
 *.DS_Store
 
 #vscode
-.vscode
+.vscode
+.Rproj.user

R/.Rprofile

@@ -0,0 +1 @@
+source("renv/activate.R")

R/basel_weather_intro.R

@@ -0,0 +1,104 @@
+library("dplyr")
+library("lubridate")
+library("magrittr")
+library("janitor")
+library("ggplot2")
+library("here")
+
+#Set the working directory
+setwd(paste0(here(),"/R"))
+
+# Read the weather data
+weather = read.csv("../data/basel_weather.csv")
+
+# Clean the column names
+#   This function removes special characters, spaces etc
+weather %<>%
+    janitor::clean_names()
+
+colnames(weather)[2] = "temperature"
+
+# Create various date variables and add them to the dataframe
+weather %<>%
+  mutate_at(vars(timestamp), as.character) %>%
+  mutate(date = as.Date(substr(timestamp, 1 ,8 ), format="%Y%m%d")) %>%
+  mutate(year = lubridate::year(date)) %>%
+  mutate(month = lubridate::month(date)) %>%
+  mutate(day = lubridate::day(date)) %>%
+  mutate(day_of_year = lubridate::yday(date))
+
+extract_hour = function(timestamp){
+  t = substr(timestamp, 10, 11)
+  as.numeric(t)
+}
+
+weather %<>%
+  mutate(hour = extract_hour(timestamp))
+
+# Plot the hourly temperature over time
+ggplot(weather) +
+  geom_line(mapping=aes(x=date, y=temperature), col="blue") + 
+  ggtitle("Basel Temperature (Hourly)") + 
+  xlab("Date") + 
+  ylab("Temperature (deg C)")
+
+# Plot the daily temperature over time
+daily_df = weather %>%
+              group_by(date, day) %>%
+              summarise(temperature = mean(temperature),
+                        day_of_year = unique(day_of_year))
+
+ggplot(daily_df) +
+  geom_line(mapping=aes(x=date, y=temperature), col="darkblue") + 
+  ggtitle("Basel Temperature (Daily)") + 
+  xlab("Date") + 
+  ylab("Temperature (deg C)")
+
+
+# Visualise the seasonality using a cyclical transformation and a spider plot.
+ggplot(daily_df) + 
+  geom_line(mapping=aes(x=day_of_year, y=temperature), col="blue") +
+  coord_polar() + 
+  ggtitle("Yearly Seasonality")
+
+# Create a function for calculating the moving average
+moving_average = function(x, k = 14){
+  x_smooth = rep(NA, length(x))
+  for (i in (k+1):length(x)){
+    x_smooth[i] = mean(x[(i - k):i])
+  }
+  return(x_smooth)
+}
+
+daily_df$temp_smooth_ma_7 = moving_average(daily_df$temperature, k=7)
+daily_df$temp_smooth_ma_14 = moving_average(daily_df$temperature, k=14)
+daily_df$temp_smooth_ma_365 = moving_average(daily_df$temperature, k=365)
+
+# Plot the Smoothed Moving Average variables
+ggplot(daily_df) + 
+  geom_line(mapping=aes(x=date, y=temperature), col="blue") + 
+  geom_line(mapping=aes(x=date, y=temp_smooth_ma_7), col="darkred")
+
+ggplot(daily_df) + 
+  geom_line(mapping=aes(x=date, y=temperature), col="blue") + 
+  geom_line(mapping=aes(x=date, y=temp_smooth_ma_14), col="darkred")
+
+ggplot(daily_df) + 
+  geom_line(mapping=aes(x=date, y=temperature), col="blue") + 
+  geom_line(mapping=aes(x=date, y=temp_smooth_ma_365), col="darkred")
+
+
+
+# Apply a gaussian filters as smoothers
+daily_df$temp_gf_30 = ksmooth(x=daily_df$date, y=daily_df$temperature, bandwidth = 60, kernel="normal")$y
+daily_df$temp_gf_90 = ksmooth(x=daily_df$date, y=daily_df$temperature, bandwidth = 180, kernel="normal")$y
+
+
+# Plot the smoothed functions
+ggplot(daily_df) + 
+  geom_line(mapping=aes(x=date, y=temperature), col="blue") + 
+  geom_line(mapping=aes(x=date, y=temp_gf_30), col="darkred")
+
+ggplot(daily_df) + 
+  geom_line(mapping=aes(x=date, y=temperature), col="blue") + 
+  geom_line(mapping=aes(x=date, y=temp_gf_90), col="darkred")