In brisneve/ggplottimeseries: Visualisation of Decomposed Time Series with ggplot2

knitr::opts_chunk$set(echo = TRUE)

Introduction

This package contains functions that allow users to plot decomposed time series data with ggplot2.

Installation

To install and load this package, you can use the following codes:

devtools:: install_github("brisneve/ggplottimeseries")
library(ggplottimeseries)

Functions

Decompose time series data into its components (i.e. trends, seasonal, random) and collate them into a data frame.

To demonstrate this, we will use the function dts1 to decompose the daily sea surface temperature data into its three components. This dataset has 2142 observations on the following variables:

1. date - a vector of time of observation (class: data, format: YYYY-MM-DD)
2. sst - sea surfae temperature data (class: numeric)

The dts1 function requires the following inputs:

1. x - dates of the observations
2. y - observations
3. z - frequency of the obserbation over a period of time (e.g. year)
4. type - type of timeseries (additive or multiplicative)

Try the codes below and explore the output data frame which contains the components of the time series:

data(sst)

x <- sst$date
y <- sst$sst
z <- 365.25 #number of days in a year

df <- dts1(x,y,z, type = "additive")

head(df)

Alternatively, if you have a time series-class data you can use the function dts2. See how this function works using the co2 dataset that contains a time series of atmospheric concentration of CO2. This function creates a dataframe of the time series components similar to that of dts1.

data(co2)

df <- dts2(co2, type ="additive")

head(df)

We now have a data frame that we can plot using ggplot. This package offers suite of functions to plot the decomposed time series data in ggplot and can be integrated with ggplot functions. We will use the co2 dataset to illustrate these functions.

Visualise the observed, trend, seasonal, and random components of time series into one figure

library(ggplot2)
library(tidyr)


#plots decomposed time series into one figure
ggdecompose(df)+
  xlab("Date")+
  ylab("Atmospheric Concentration of CO2")

Visualise components of time series into separate figures

#plots the observations 
ggobserve(df)+
  xlab("Date")+
  ylab("Observed Atmospheric Concentration of CO2")

#plots the trend 
ggtrend(df)+
  xlab("Date")+
  ylab("Trend of Atmospheric Concentration of CO2")

#plots the seasonality 
ggseason(df)+
  xlab("Date")+
  ylab("Seasonality of Atmospheric Concentration of CO2")

#plots the random variation 
ggrandom(df)+
  xlab("Date")+
  ylab("Random Variation of Atmospheric Concentration of CO2")

Detrend and de-seasonalise timeseries and visualise the data

ggdetrend(df)+
  xlab("Date")+
  ylab("Detrended: Concentration of Atmospheric Concentration of CO2")

ggdeseason(df)+
  xlab("Date")+
  ylab("De-seasonalised: Concentration of Atmospheric Concentration of CO2")

Citation:

Brisneve Edullantes (2019). ggplottimeseries: Visualisation of Decomposed Time Series with ggplot2. R package version 0.1.0.

brisneve/ggplottimeseries documentation built on May 7, 2019, 3:08 p.m.