In wzb-ipi/dashdash: Make simple dashboards for live data display

knitr::opts_chunk$set(echo = TRUE)
library(dashdash)

1. Motivation

Lots of Covid-19 studies are going on. This tool can be used to present basic results in a simple manner in real time (or, as quickly as you can make your data available).

Here's an example from Sierra Leone: https://sl-dashboard.github.io/corona/

Here's another one from Nigeria: https://wzb-ipi.github.io/nigeria_dashboard/

Besides helping you display data we have two ulterior motives:

• If teams use this it would be easy to use it to create a meta-dashboard across teams and countries making it easy to locate and view the good work being done
• If teams use this then it would not be hard to pool studies together to calculate cross study aggregates. In addition to the basic information, good aggregation would require sampling weights information.

Notes:

• Software: the dashboard is produced using Rmarkdown but the inputs are just data sets.

• Modifying the software. You can copy or fork the repository to modify it to suit your needs (pull requests also welcome!). The package contains very simple code. The main function is just a call to rmarkdown::render to compile an .Rmd file. The .Rmd file pulls in some "child" .Rmds. These .Rmds determine the overall stucture of the dashboard. For content, the Rmds call to simple functions (e.g. plot_aggregates()) that mostly implement calls to dplyr and ggplot to make tables and graphs. So edit the .Rmds (found in \inst) to modify the structure and edit the .Rs (found in \R) to modify the analyses.

• Hosting: If you can get the input files to us and can regularly upload your data to Github we think we can produce and host the dashboard on this site and catalogue it with other ones. Alternatively the package produces a html file that you can host anywhere. If you have a Covid survey that you want to connect with this drop a note to [e-mail] or [e-mail].

• Note on open source material: All material here is open source. Data posted here will be made public and should not contain any personally identifiable information.

2. Inputs

The dashboard function requires three key inputs:

my_vars dataframe

A spreadsheet that says what your variables are and to which families they belong. For example:

 my_vars <- data.frame(
   variable = c("market_open", "price_rice", "aware", "water"),
   family = c("Markets", "Markets", "Actions", "Actions"),
   short_label = c("Is market open?", "Price of a rice", "Aware of Covid19", "Access to water"),
   description = c("details on market open", "Price of a cup of rice", "Details on aware of Covid19", "Details on access to water"),
   min = c(-2, -1, 0, NA),
   max = c(2, 1, 1, NA),
   stringsAsFactors = FALSE
 )

kable(my_vars, caption = "sample `my_vars` dataframe")

Good visualization requires that short labels are actually short and that you provide the range of the variable (min and mex values). The latter point, however, is optional. To tackle problems with the former we provide a tab with definition measures.

my_data dataframe

Your data, cleaned and transformed to the point where it is ready for display:

my_data <-
   expand_grid(id = c("Bo", "Bombali", "Bonthe"), date = as.Date(c("2020-06-13", "2020-06-14", "2020-06-15", "2020-06-16", "2020-06-17", "2020-06-18", "2020-06-19"))) %>%
   data.frame(stringsAsFactors = FALSE) %>%
     mutate(n = 3) %>%
     uncount(n) %>%
     dplyr::mutate(
     market_open     = rbinom(n(), 1, .5),
     price_rice  = rbinom(n(), 1, .5),
     aware   = rbinom(n(), 1, .5),
     water   = rbinom(n(), 1, .5)) %>%
     distinct()

datatable(my_data, caption = "sample `my_data` dataframe")

The key requirements are that the dataset contains:

• an id variable which gives the level at which you want the data to be aggregated for averages -- this variable should correspond to your id variable in your shape files.
• a date variable whch is a string variable of the form YYYY-MM-DD

my_args dataframe

Third, a spreadsheet with dashboard specific arguments, inluding any text you want to appear. You can provide title and author information in here, or you can provide these directly to the dashboard function.

my_args <- data.frame(
  intro_text = "Intro text",
  intro_note = "Lots of great people worked on this. More information here: ...",
  data_note = "Study specific note about data source",
  map_path = paste0(getwd(), '/shapefiles'),
  map_layer = "SLE_adm3",
  map_region =  "NAME_2",
  group = "District",
  scale_vars = TRUE,
  stringsAsFactors = FALSE
)

my_args %>% kable(caption = "sample `my_args` dataframe")

3 Produce Dashboard

The dashboard is then produced by dashdash::dashdash like this:

setwd("~/Documents/GitHub/dashdash/docs")
wd <- getwd()

dashdash::dashdash(
  output_file = paste0(wd, "/example.html"),
  title = "A sample dashboard for my study",
  subtitle = "What's special about this study",
  my_data = my_data,
  my_vars = my_vars,
  my_args = my_args)

Note that output_file, my_data, my_vars and my_args are the only required inputs. In this example we also provide title and subtitle directly, but these could also be contained in my_args.

remotes::install_github("wzb-ipi/dashdash")

dashdash::dashdash(
  output_file = "example.html",
  title = "A sample dashboard for my study",
  subtitle = "What's special about this study",
  my_data = my_data,
  my_vars = my_vars,
  my_args = my_args)

You can then view it: example.html

4. Extensions

Given a lot of feedback we have received with teams engaging with our package we added new features.

Daily data or moving average with argument trend

Say that you data is coming with a relative sparse frequency. If you pass trend=list("moving_average")you can get a three day moving average. Also you could passtrend=list("daily"," moving_average") to get both type of plots.

my_blog dataframe

Now imagine that you have found some really interesting trend in your data, and you feel the urge to make explicit this preliminary analysis. You could do so by passing on a dataframe to the argument my_blog

my_blog <- data.frame(
   family = c("Markets", "Actions"),
   entry_aggregates = c("We can see a very interesting trend in the variables under Markets tab", "In the Actions tab we can also see a very interesting across districts"),
   entry_disaggregates = c("When looking at disaggregates plots in Markets we found more interesting things", "Now disaggregate plots are also interesting for Actions tab"), stringsAsFactors = FALSE
 )

datatable(my_blog, caption = "sample `my_blog` dataframe")

Try running our extended dashboard now!

setwd("~/Documents/GitHub/dashdash/docs")
wd <- getwd()

wd <- getwd()
dashdash::dashdash(
  output_file = paste0(wd, "/example.html"),
  title = "A sample dashboard for my study",
  subtitle = "What's special about this study",
  my_data = my_data,
  my_vars = my_vars,
  my_args = my_args,
  my_blog = my_blog,
  trend = list("moving_average", "daily")
)

5. Aggregating dashboards

Coming next

wzb-ipi/dashdash documentation built on Aug. 30, 2020, 4:42 p.m.