01-intro.Rmd

---
chapter: 1
knit: "bookdown::render_book"
---


# Introduction {#intro1}

Maps can contribute to the interpretation of spatial distributions of disease occurrence, and help to locate disease clusters. Disease data is commonly aggregated to political areas; privacy is one reason for this and another is that it is the responsibility of the political entity to respond. The typical visualisation for aggregated spatial data is a choropleth map, where areas are coloured according to a numerical value. A choropleth map is the most common display for the presentation of disease data.

Choropleth maps can mislead the map reader, as the attention of the map user is distributed according to the size of the area. Australia provides an extreme example of potential bias and loss of valuable information. In Australia, the geography mismatches the population, because the communities are densely populated in the inner city areas, especially around the capital cities, and the coastline. There are alternative visualisation methods, like cartograms, that have been developed to correctly focus on the population dense areas. These alternatives should be considered when planning the communication of geospatial statistics, as visualisations should be chosen to best represent the spatial distribution. 

This thesis research is motivated by the Australian Cancer Atlas, which presents the spatial patterns of cancer in Australia. The aim of this thesis is to contribute an algorithm that creates effective visualisations for the communication of geospatial population statistics.


## The Australian Cancer Atlas

The Australian Cancer Atlas (ACA) is an online, interactive web tool for exploring the impact of cancer on Australian communities. The prominent display used by the ACA to present incidence rates or excess death rates is the choropleth map. The set of geographic units shown are the Australian Statistical Areas, at Level 2 (SA2s). There are almost 2,200 individual SA2s.

The choropleth map used in the ACA is familiar to the general public of Australia. It is appropriate to use this map display as users can orient themselves on the map and find the geographic areas relevant to them.
However, when the intention of the map user is to understand the whole spatial distribution, the information derived from the colours displayed on the map can be misleading.
The rural areas in outback Australia are over emphasised, and the densely populated inner city areas are not given enough attention, as they cannot be seen using a choropleth map.

## Testing graphical displays

Visual inference testing will be used to determine if the communication of population geospatial statistics is more effective when using the new alternative hexagon tile map display.
Buja et al [@GIIV] provide the 'lineup' protocol as a formal framework for testing visual statistical methods. The new alternative visualisation method can be tested by implementing this framework.
It takes inspiration from a police lineup.
The lineup protocol can be used to test if the hexagon tile map is effective, a map displaying a real population based distribution can be hidden in a collection of maps that display null distributions [@chowd]. 


<!-- Talk about same data, same info for lineup, power comparison of competing designs -->


## Aims and Objectives {#sec:aims}

This thesis aims to provide a solution to presenting geospatial data regarding populations. 
It considers the visualisation methods developed over the past two centuries that shift the focus from the geographic map base.

1. *Devising an algorithm for creating hexagon tile maps of Australia:* The algorithm will take geospatial areas and create an alternative visualisation of the spatial distribution.

2. *Test the effectiveness of the hexagon tile map relative to the choropleth map:* The hexagon tile map produced by the algorithm will be contrasted with the traditional choropleth map, applying the same colour methods to represent the data. 

3. *Communicating the relationship between the hexmap and choropleth map through animation:* Animations can maximise the benefits of both visualisation methods when communicating to the public. The use of animations may control how people follow a recognisable map of Australia into an alternative visualisation for inference. 

## Research Contributions

This research contributes a new algorithm for creating hexagon tile map displays. It contributes an R [@R] package which implements the algorithm and allows R users to create their own visualisations.
It presents a case study that contributes to a growing field of visual inference studies. Applying the lineup protocol to spatial data by comparing a choropleth map to a hexagon tile map display.
It also shows how it can be used in practice to effectively communicate population related cancer distributions.


## Thesis Structure

The thesis is structured as follows: Chapter two contains a literature review. 
The literature reviews considers the current peer reviewed literature and published books that explore spatial distributions of cancer across the globe.
It also considers how to evaluate the visualisation methods used for spatial data.

Chapter 3 explores the algorithm to create hexagon tile maps and the code used to create a small example of Tasmania in Australia.
Chapter 4 discusses an experiment that uses visual inference. It contains the methods and results that compare the use of a choropleth map and a hexagon tile map.
Chapter 5 summarises the contributions of this thesis and possible future work.