In ukgovdatascience/orgsurveyr: Visualising and Analysing Organisation Data As Tidy Networks

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-"
)
library(tidygraph, quietly = TRUE)
library(ggraph, quietly = TRUE)
library(dplyr, quietly = TRUE)
library(orgsurveyr, quietly = TRUE)

orgsurveyr

The goal of orgsurveyr is to facilitate the use of Thomas Lin Pedersen's excellent tidygraph and ggraph packages with organisational information used by many Human Resources departments, in particular employee surveys.

Installation

You can install orgsurveyr from github with:

# install.packages("devtools")
devtools::install_github("ukgovdatascience/orgsurveyr", build_vignettes = TRUE)

There is also infrastructure within the repo to install and run orgsurveyr within Docker. See the section at the end of this help file for details but essentially you clone the repo and type make run. More help is available with make help.

Example - ggraph/tidygraph

An organisation is a very basic type of network known as a tree. The tidygraph package lets us simulate and represent a tree structure as follows:

tg <- tidygraph::create_tree(13,3) 
tg

The ggraph package can be used to plot networks with the familiar ggplot2 syntax. Below the simulated organisation is plotted as a dendrogram:

  ggraph(tg, 'dendrogram') + geom_edge_diagonal() + 
  geom_node_point(size=5) + theme_bw()

Example - orgsurveyr

The orgsurveyr package provides detailed vignettes, convenience functions, shiny apps and example data to help HR analysts make use of the ggraph and tidygraph packages in the analysis of organisation data.

A realistic organisation can be simulated and plotted:

set.seed(1236)
tg2 <- create_realistic_org(4,3, prob=0.3)
plot_org(tg2)

Individual level data also be simulated first by simulating the number of individuals in unit:

tg3 <- simulate_unit_size(tg2)
tg3

Followed by simulating the individuals and any additional variables. The commands can be piped as per tidyverse convention:

tg3_individuals_df <- tg3 %>% 
  simulate_individuals_df() %>%
  mutate(test_var2 = purrr::map_dbl(individual_id, ~rnorm(1, 20,3)))
tg3_individuals_df

The individual data is aggregated at the unit level. This can be either be just those individuals associated with an organisational unit or (more usefully) a cumulative aggregation involving all individuals that are part of that unit and any units below it in the organisation:

summary_df <- calc_summary_df(tg=tg3, df=tg3_individuals_df, 
                              selected_vars=c('test_var', 'test_var2'), is_cumulative=TRUE)
summary_df

The aggregated data can be plotted on the dendrogram:

plot_org(tg3, fill_var = 'test_var', summary_df)

Larger organisations can be represented as starburst plots:

set.seed(1236)
tg4 <- create_realistic_org(5,5, prob=0.3) %>%
  simulate_unit_size()

tg4_individuals_df <- tg4 %>%
  simulate_individuals_df() %>%
  mutate(test_var2 = purrr::map_dbl(individual_id, ~rnorm(1, 20,3)))

tg4_summary_df <- calc_summary_df(tg=tg4, df=tg4_individuals_df, 
                              selected_vars=c('test_var', 'test_var2'), is_cumulative=TRUE)

plot_org(tg4, fill_var = 'test_var', df = tg4_summary_df, is_circular = TRUE, is_dendrogram = FALSE) +
  scale_fill_gradientn(colours=RColorBrewer::brewer.pal(11, 'PiYG'))

An interactive visualisation in shiny can be launched:

orgviz(tg=tg3, df=summary_df)

Docker

The git repository contains infrastructure for installing and running orgsurveyr within Docker. This is useful if you quickly and easily want to try out the package before installing it, and is also useful for trying out the shiny apps.

First clone the git repository to your local machine:

git clone https://github.com/ukgovdatascience/orgsurveyr.git

Then from within the repo type make run. This will:

• Build the Docker image
• Start a container
• Open RStudio and Shiny server in your default browser (localhost:8789 and localhost:3839 respectively)

To stop and remove the Docker container type make remove. More help is available by typing make help.

Further reading/viewing

• Data Imaginist - Thomas Lin Pedersen's blog
• 'Tidying up your network analysis with tidygraph and ggraph': RStudio::conf presentation from Thomas Lin Pedersen
• useR 2018 Interview with Thomas Lin Pedersen

ukgovdatascience/orgsurveyr documentation built on May 4, 2019, 7:41 p.m.