Skip to main content

EDAV Fall 2021 Tues/Thurs Community Contributions

55 An Introduction to Dygraph in R

Yanyun Chen

55.0.1 These are the codes demonstrated in the video tutorial. The video tutorial can be found at https://www.youtube.com/watch?v=ewQdzu47yZs

Data Source: FRED, https://fred.stlouisfed.org/categories/9?t=cpi&ob=pv&od=desc#

55.0.2 Project Description

The main motivation of this project is the ubiquity of time series data. One thing I noticed in my past internship and project experience is that the need for exploratory analysis and visualization of such data is high, but people are less familiar with it than other types of data. Moreover, the r package dygraphs provides much easier and intuitive functionality to draw time series data when compared to ggplot or plotly (which were introduced in class), and it additionally has interactive features that are useful for long-term time series data. I personally think the video tutorial and example codes are organized and rendered in a clean and friendly way that anyone with some r background would understand and learn. During the making process, I’ve also benefitted a lot from exploring this package, especially plotting prediction range and actual data in such a quick and easy way. However, what I might do differently next time is to pay more attention to the best practices of visualizing time series data, such as not filling the area.

55.0.3 The Basic 

cpi_data <- read_excel("resources/intro_to_dygraph/cpi.xlsx", col_types = c("date", 
    "numeric", "numeric"))
colnames(cpi_data) <- c("date","us","china")
us_cpi <- ts(cpi_data$us, frequency=12, start=c(1999,9))
dygraph(us_cpi)

Note: In the video, fill is demonstrated. But filling time series graph is generally not recommended.

dygraph(us_cpi, main = "CPI, not seasonally adjusted: All Items for United States") %>% 
  dySeries("V1", label = "United States CPI",color='red') %>%
  dyAxis("y", label = "Index (2015=100)", valueRange = c(50, 120)) %>%
  dyOptions(axisLineWidth = 1.5, 
            drawGrid = FALSE) %>%
  dyRangeSelector(height = 20) %>%
  dyLegend(show = "follow")

55.0.4 Multiple Series 

cpi <- ts(cpi_data[,-1], frequency=12, start=c(1999,9))
dygraph(cpi)
dygraph(cpi,main="CPI, not Seasonally Adjusted: All Items") %>%
  dySeries("us", label = "United States",color='red') %>%
  dySeries("china", label = "China",color='blue') %>%
  dyAxis("y", label = "Index (2015=100)") %>%
  dyOptions(axisLineWidth = 1.5, drawGrid = TRUE) %>%
  dyRangeSelector(height = 20)

55.0.5 Events

Note: In the video, drawing event line along with shaded region is demonstrated. Another option is to draw annotation along with shaded region, which is demonstrated below.

dygraph(cpi, main="CPI, not Seasonally Adjusted: All Items") %>%
  dySeries("us", label = "United States") %>%
  dySeries("china", label = "China") %>%
  dyRangeSelector(height = 20) %>%
  dyEvent("2019-12-01", "COVID-19", labelLoc = "bottom") %>%
  dyAnnotation("2008-05-01", "2008 Financial Crisis", attachAtBottom = TRUE, width = 150) %>%
  #dyEvent("2008-05-01", "2008 Financial Crisis", labelLoc = "bottom") %>%
  dyShading(from = "2008-01-01", to = "2008-12-01")

55.0.6 Prediction 

china_cpi <- ts(cpi_data$china, frequency=12, start=c(1999,9))
hw <- HoltWinters(china_cpi)
predicted <- predict(hw, n.ahead = 36, prediction.interval = TRUE)
all <- cbind(china_cpi, predicted)

dygraph(all, "CPI, not Seasonally Adjusted: All Items for China") %>%
  dySeries("china_cpi", label = "Actual") %>%
  dySeries(c("predicted.lwr", "predicted.fit", "predicted.upr"), label = "Predicted")
china_cpi_200 <- ts(cpi_data$china[1:200], frequency=12, start=c(1999,9))
hw <- HoltWinters(china_cpi_200)
predicted <- predict(hw, n.ahead = 64, prediction.interval = TRUE)
all <- cbind(china_cpi, predicted)

dygraph(all, "CPI, not Seasonally Adjusted: All Items for China") %>%
  dySeries("china_cpi", label = "Actual") %>%
  dySeries(c("predicted.lwr", "predicted.fit", "predicted.upr"), label = "Predicted")