DJM Topic 3 - Visualising Proportions

VIsualising Proportions

what are proportions

relative sizes of a part compared to the whole often expressed in percentages

what are proportions grouped by

categories or subcategories

most important for proportions

maximum, mimimum, and overall distribution

ranking analysis?

sorting the values

part-to-whole analysis

expressing the values as percentage of the whole

what are the porportion data patterns

uniform

uniformly different

non uniformly different

increasingly different

decreasingly different

alternating differences

exceptional

uniform

all values are roughly the same

uniformly different

differences from one value to the next differ by about the same amount

non uniformly different

differences from one value to the next vary significantly

increasingly different

differences from one value to the next increase

decreasingly different

differences from one value to the next decrease

alternating differences

differences from one value to the next can be very small at first then suddenly very large then back to small again

exceptional

one or more value are extraordinarily different from the rest

which patterns are interesting?

unusual differences from one value to next

setss of values appear to be grouped

significant breaks in a pattern

obvious exceptions from the norm

proportion data displays

pie chart, donut chart, stacked area chart, stacked bar chart, treemap, pareto chart

pie chart

values altogether add up to 100%

use 3 or less values

donut chart

same as pie chart with hole in middle

when to use pie chart?

do the values make up a meaningful whole

are the values mutually exclusive

do u want to compare values with each other or values to the whole

how many variables do u want to show

stacked area chart

used to show changes over time for several variables

stacked bar chart

used for a few distinct points in time

treemap

uses area size of rectangles to show relative proportion and works well if there is hierarchical structure with parent and child nodes

quantities of a variable is shown by area size

2nd measure can be colour/gradient of the rectangles

pareto chart

used to show cumulative contributions as parts to whole

can be made up using column and line chart

cautions on pareto chart

wildly fluctuating system will produce inconsistent pareto rankings

repeated pareto analyses will help confirm rankings but best way is to use control chart to ensure system is stable

when to use pareto

can data be grouped in categories

is ranking of categories important

used when finding out frequency of problems in a process

techniques and best practices

grouping categorical data in an ad hoc manner

using pareto chart with percentile sclaes

re expressing values to solve quantitative scaling problems

using line chart to show ranking changes through time