unit 1 info vis

2 purposes of info vis

analysis & presentation

analysis

to better understand ur data and act upon that understanding

presentation

to communicate and inform others more effectively

what does EDA stand for?

exploratory data analysis

when to use EDA

1. u dont know exactly what ur looking for

2. u dont have any prior questions

3. u want to know what to questions to ask

4. u want to get a better feel for the data

key components of info vis

representation & interaction

representation

chart type

interaction

allowing people to explore (zoom, scroll)

data attributes

columns

data items

rows

basic types of attributes

nominal, ordinal, interval

nominal

2+ categories with specific no order

ordinal

2+ categories that are ordered or on a spectrum

interval

continous numeric (e.g. age)

User tasks

retrieve value, filter, compute derived value, find extremum, sort/rank, determine range, characterize distribution, find anomolies, cluster, correlate

how to compute derived value

given set of data cases, compute total numeric representation of those data cases (aka like avg.)

how to find extremum

find data cases possessing an extreme value of an attribute over its range w/in the data set

what histogram distribution is this?

normal

what histogram distribution is this?

right skewed

what histogram distribution is this?

left skewed

what histogram distribution is this?

uniform

what histogram distribution is this?

bimodal

what histogram distribution is this?

multimodal

what correlation is this

positive

what correlation is this

negative

what correlation is this

no relationship

what correlation is this

curvilinear

how to interpret correlation of scatter plots?

There exists a [strength], [type] association btwn _____ and ______, such that [x] tend to be [y]

Munzner taxonomy

framework of visualization tasks as pairs of actions & targets

Stasko taxonomy

low-level components of analytical activity in info vis (focused on intents & goals ppl have, rather than the UI) - user tasks

how did stasko’s taxonomy come to be

affinity diagramming

what is affinity diagramming?

visual tool that helps u organize info by sorting ideas into diff groups based on their relationship to one another

what makes up actions?

analyze, search, query

what makes up targets?

all data, attributes, network data, spatial data

what is the curse of knowledge

the concept that explains why once you know something, it becomes hard to imagine not knowing it. can cause you to assume the reader has the same knowledge and expertise as u do

mental models

mental constructs that are created based on our experience IRL

reader-driven

overall less messaging and more open interactivity (flexible)

author-driven

order matters, stronger messaging and interactivity can be limited

data story structures

martini glass, interactive slideshow, drill-down story

how many different visual encodings in one visualization?

one for ea attribute

what is this

parallel set

what is this

mosaic plot

what is this

stacked barchart

what is this

star plot

what is this

bubble chart

what is this

tile map

what is this

heat map

Edward Tufte

visualizations are a way to explain fancy stats so avg. ppl can understand (graphical excellence)

Nigel Holmes

chart junk isn’t always bad, and embellishments can help enhace engagement

graphical excellence

communicate essence and substance of the data & stats through a chart in a precise and efficient way

graphical integrity

tell the truth about data and do it without diminishing the aethetic experience

Lie factor

size of effect shown in graphic / size of effect in data

data-ink-ratio

ink used to draw data / ink used in graphic

Holme’s 2 tips

declutter & focus

uncertanties

used to describe 3 things, accurracies, precision and reliability

reliability

how likely u r to get the same results thru repetition

visualizing uncertainty GOOD

gradient or violin plot

visualizing uncertainty BAD

bar chart w/ error bars

effective presentation of uncertainty

as a set of discrete, countable outcome

hypothetical outcome plots

presenting uncertainty as a sample over time, where ea sample is a new frame in an animated visual

plinko dot plots

looks like the plinko game, designed to approvimate the data-generation process

how to evaluate uncertainty visualizations?

1. does it support the task?

2. does it match intuitions?

precision range

COSP (color, orientation, size, position)

simpsons paradox

trends that appear individually in subgroups of a dataset reverse when the data are aggregated and looked at as a whole