This post follows the analysis of Taiji Cove drive hunt. You must first read that analysis before you proceed here.
Graph objective
The graph objective is to document the effect of “Relative number of captured to killed cetaceans” at the Taiji Cove hunting season. The graph objective follows an initial investigation in the Taiji Cove drive hunt analysis, where it was discovered that there is an increase trend in the rate of cetaceans captured alive, as capturing and training dolphins to be sold at dolphinariums is more lucrative than slaughtering the animals for their meat.
Data management
The data source is described in the Taiji Cove drive hunt page. The data reports the number of cetaceans killed and numbers captured to be sold in dolphinariums.
The question of rate of change will be analysed using indices with baseline year 2001 (the earliest years in the dataset).
The Stata code describing the data management is provided at the end of this page.
Visual implantations
Although the question is related to a timeline evolution I choose to avoid working with timeline plots because of the suggested comparison of two indices. That is, if I were to employ timeline plots then the timeline of those killed will be decreasing and those captured will be increasing much faster and they crossing of the two lines would wrongly suggest that there are more captured animals than killed animals.
Instead, I choose to form these comparisons using the area visual implantation, and specifically horizontal bars that will be shown to deviate away from the baseline value of 1 (the index baseline).
In terms of encoding bars in Stata, I have always found the canned solution of bar plot commands to be somewhat cumbersome to work with, so I switch to paired-coordinate plot and give the impression of bars by encoding lines of different width. In other words, I emulate the area implantation by using thick line implantations.
Retinal variables
The graph objective requires the contrast of two categories: rate of change killed and rate of change number captured. I choose to contrast the two categories with the colour retinal variable using low saturation navy and dark orange colours. I assign the brightest colour of the two to the numbers captured in order to draw more attention to the graph objective.
In addition, as discussed just above, I use the size retinal variable to decrease the area of ‘killed’ to half as large as that of ‘captured’, because of two reasons: (i) the focus is on cetaceans captured, (ii) I need a way to show the information from the overlapping indices in certain years.
The orientation retinal variable also plays an important role as the leftwards direction of bars encodes decreases relative to the baseline year 2001, and the rightwards direction encodes increases relative to 2001.
Graph identification
Instead of showing a legend, I internally identify the encoding choices in the graph. In navy colour I write “Killed index” and in orange colour I write “Captured index”.
Given the small dataset, I directly identify the data values next to the height of each bar using the associate colours. I also identify in dark gray colour the value of 1, which is the baseline index value as at year 2001, plus I encode directional arrows with the leftwards arrow describing decreases and the rightwards arrow describing increases.
External identification adds a graph title describing the graph objective, and a note that acknowledges the source of the data. I also identify the year labels on the y-axis.
Graph enhancement
A key graph enhancement step is the encoding of reference areas that assist the decoding of alternating years. I emulate reference areas by increasing the thickness of grid lines and reducing their color value. This helps decoding by grouping information within boxes areas (this is based on the Gestalt principles of visual perception).
The use of direct identification, as discussed just above, obviates the need of an x-axis scale so this is suppressed altogether.
I reduce the overall visual prominence of the y-axis year labels and eliminate all axes ticks. Finally, I adjust the aspect ratio to 1.5:1 given the vertical orientation of the information.
Visual decoding/perception
Here is my proposed solution:
This an attempt to encode a contrast between two indices, i.e. measures of rate of change with common baseline of 2001.
The data graph makes is clear that there is a disproportionate increase over time in the rate of captured cetaceans by comparison to the killed cetaceans over time. As activists focused their attention towards reducing the number of animals killed, it seems that the fisherman adapted by directing more attention to capturing and selling live animals.
Download the Stata code for reproducing this analysis: taiji_cove_killed.do
Graph workflow 