This is the script for the htmlwidgets for R demo at the R-Ladies Coding Club’s May Event. Published doc available on RPubs: http://rpubs.com/crt34/widgetdemo

Demo of selected htmlwidgets

– A. Time-Series Plot
– B. Heatmap
– C. 3D Scatterplot
– D. 3D Globe

What are htmlwidgets for R?

http://www.htmlwidgets.org

A growing number of htmlwidget R libraries have been developed from popular JavaScript charting libraries, such as D3.js, plotly.js, Leaflet.js etc.

This htmlwidgets framework/interface therefore allows R users to generate interactive JavaScript data visualisations using R code, through the normal operation of loading relevant libraries and calling appropriate functions.

These interactive plots can be displayed in RStudio, embedded in R markdown documents and Shiny web apps, and shared in standalone html files over email, dropbox etc.

What Use-Cases do we have for htmlwidgets for R?

– Data Storytelling: presenting & communicating findings
– Exploratory Data Analysis (EDA)

Install Packages for all Demos

install.packages("dygraphs")
install.packages("xts")
install.packages("d3heatmaps")
install.packages("threejs")

A. Time-Series Plot

Produce an interactive time-series visualisation using the dygraphs R library. Objects for plotting must be in xts format (extensible time series).

Time-series data = Australian wine sales in no. of bottles, by wine makers, between Jan 1980 - Aug 1994.

Using the stl function to perform the seasonal trend decomposition, which extracts the seasonal component, trend, component and remainder from the original data.

(i) Call dygraph & xts libraries

library(dygraphs)
library(xts)

(ii) Load Data

In this instance the Australian Wine dataset wineind being used is from the forecast package so we load this first, then call the data.

(Original source: http://data.is/TSDLdemo)

library(forecast)
data("wineind")

(iii) Explore Data

?wineind
class(wineind)
wineind
plot(wineind) #base R plot

(iv) Apply Decomposition algorithm

Apply stl algorithm, which divides the time series into 3 components, trend, seasonality & remainder, using Loess (a method for estimating nonlinear relationships).

stl documentation: https://stat.ethz.ch/R-manual/R-devel/library/stats/html/stl.html

stl_wine <- stl(wineind, s.window = "periodic") #s.window controls variation of seasonal component
stl_wine
plot(stl_wine) #base R plot

(v) Prepare decomposed data as a single xts object

First, convert each decomposed component into a separate ts (time series) object, and then into xts (extensible time series) object:

#(1980, 1) is start year & unit of observations, 12 is the no. of observations per period i.e. monthly

random_stl <- ts(stl_wine$time.series[,3], start = c(1980, 1), frequency=12)
random_stl.xts <- as.xts(random_stl)

seasonal_stl <- ts(stl_wine$time.series[,1], start = c(1980, 1), frequency=12)
seasonal_stl.xts <- as.xts(seasonal_stl)

trend_stl <- ts(stl_wine$time.series[,2], start = c(1980, 1), frequency=12)
trend_stl.xts <- as.xts(trend_stl)

Secondly, column bind the 3 xts components together to create a single xts object for plotting:

wine.plot <- cbind(random_stl.xts, seasonal_stl.xts, trend_stl.xts)

#re-name columns
colnames(wine.plot) <- c("remainder", "seasonal", "trend")

(vi) Create Dygraph

Basic:

dygraph(wine.plot) %>%
    dyOptions(stackedGraph = TRUE)

(vii) Customise Dygraph

Add Title & Range Selector & change colours:

dygraph(wine.plot, main = "Australian Wine Sales") %>%
    dyOptions(stackedGraph = TRUE, colors=RColorBrewer::brewer.pal(3, "Set1"))%>%
      dyRangeSelector()

B. Heatmap

Produce a D3 heatmap using the d3heatmap R library.

(i) Call d3heatmap library

library(d3heatmap)

(ii) Load Data

Data created from personal burger blog, in wide format necessary to use this library. (https://burgerite.blogspot.co.uk)

Data available here: https://github.com/RLadiesCodingLondon/MayEvent-Chiin-htmlwidgets

burger <- read.csv("burger.csv", header = TRUE, row.names = 1)

(iii) Explore Data

5 attributes, 25 burgers reviewed.

View(burger)

(iv) Create Heatmap

Blue = higher score, Red = lower score

d3heatmap(burger)

C. 3D Scatterplot

Produce a 3D scatterplot using the threejs R library using same burger data.

(i) Call threejs library

library(threejs)

(ii) Create Scatterplot

scatterplot3js(burger$proportions, burger$ingredients, burger$patty, flip.y=FALSE)
#flip.y to control direction of y-axis

(iii) Customise Scatterplot

label <- c("proportions", "patty", "ingredients") #define axis labels

scatterplot3js(burger$proportions, burger$ingredients, burger$patty, 
               axisLabels = label,
               labels=row.names(burger),
               flip.y = FALSE,
               color=rainbow(length(burger$value)),
               stroke = NULL,
               size=burger$value/5,
               renderer = "canvas")

D. 3D Globe

Produce a 3D globe, again using the threejs R library.

(i) Load Data

Geo-location data from Google Analytics of burger blog traffic.

Data available here: https://github.com/RLadiesCodingLondon/MayEvent-Chiin-htmlwidgets

views <- read.csv("globejs.csv", header = TRUE, row.names = 1)

(ii) Explore Data

169 cities with cumulative blog page views >= 2 since Sep ’15.

View(views)

(iii) Create Globe

globejs(lat = views$Lat, long = views$Long, value = views$PageViews)

(iv) Customise Globe

traffic <- 10000*views$PageViews/max(views$PageViews)

globejs(lat = views$Lat, long = views$Long, value = traffic,
        atmosphere = TRUE,
        pointsize = 0.5)