Creating Interactive Time Series Charts With Plotly in R

April 19, 2020

When analyzing time series data in R, I often draw on a small set of helper functions to quickly visualize information. These use the excellent plotly package to create interactive charts. I thought others might find these useful as well and decided to put the code up here. This post provides a quick rundown of how these functions work.

First, grab the file plotly_chart_helpers.R from here, then load it alongside the required packages:

library("xts")
library("plotly")
library("lubridate")
source("plotly_chart_helpers.R")

Interactive line plot

To create an interactive line plot, let's first generate some sample data. Note that we need to turn our data into an xts object for it to work with our helper functions.

dat <- data.frame("a" = sin(seq(1, 9*pi, length.out = 120)) +
                    rnorm(120, 0, 50),
                  "b" = sin(seq(1, 8*pi, length.out = 120)) +
                    rnorm(120, 0, 20),
                  "c" = sin(seq(1, 7*pi, length.out = 120)) +
                    rnorm(120, 0, 50))
dat <- cumsum(dat) + seq(10000, by = 10, length.out = 120)
dat <- xts::xts(x = dat,
                order.by = seq(as.Date("2000-01-01"),
                               by = "month",
                               length.out = 120))

Let's now call the plotly_line_growth() function with our new data set:

plotly_line_growth(dat, x_lab = "Date", y_lab = "Unit")

Let's briefly look at what happens in the background. The function takes our xts object (and, optionally, a second xts object of different frequency) and calculates quarterly and annual growth rates. These calculations can be disabled by setting the parameters add_q and add_y to FALSE. By default, percentage growth rates are calculated, but if you prefer simple differences, set the parameter rate to FALSE. The function then merges these new data series with the growth rates into the main set, appending _q and _y suffixes to the respective column names. While doing this, it keeps track of the supplied color palette (the included default palette can be overridden using the colpal parameter) to ensure the colors for growth rates match those for the level series.

It then creates a plotly object, adding each series with its appropriate color while ensuring that series showing growth rates are initially set to invisible. This allows us to add buttons to the chart that selectively show and hide certain data series: a “Q” button that, when toggled, will show quarterly growth rates while hiding annual growth rates and level data, a “Y” button that does the same for annual growth rates, and an “L” button that does the same for level data.

Interactive Bar Plot

Now, let's explore the equivalent function for bar plots. This function simplifies plotting data where several series contribute to a total, such as GDP and its expenditure components, or total exports split by destination. To demonstrate this, let's add a “total” column to our sample data set. Note that the total must be the first column of the data set.

dat$total <- rowSums(dat[, c("a", "b", "c")])
dat <- dat[, c("total", "a", "b", "c")]

Next, let's use the plotly_line_bar_growth() function with this updated data set:

plotly_line_bar_growth(dat, x_lab = "Date", y_lab = "Unit", add_others = FALSE)

Internally, this function is very similar to the previous one, with one key difference: Instead of calculating growth rates, it calculates the contribution to growth for each component. For the contributions to add to the change in the total, we need to ensure that our series add up. With real world data this is often not the case, usually for methodological reasons or due to rounding. In those cases, the function will automatically add a category “others” unless told not to (by setting add_others to FALSE). Another difference is that this function adds some custom hover text which shows the percentage share of each component on mouseover.

I mostly use this code to construct dashboards that show charts with daily, weekly, monthly, or quarterly frequency data, but it can be used with lower frequencies (annual or semi-annual) by setting add_q to FALSE.

Stefan Angrick