You have received the request to upgrade the web serv

CHAPTER

•

•· The real voyage of discovery consists not in seeking new ,, landscapes but in having new eyes .

CHAPTER OUTLINE 16.1 Introduction

16.2 Tasks in Data Visualization

16.3 Visualization by Data Type

16.4 Challenges for Data Visualization

M arcel Proust

551

552 Chapter 16 Data Visua lization

16. 1 Introduction

Today's users are routinely engaging with larger and more complex volumes of data than ever before-and not just for professional situations as part of their jobs but also for personal and recreation purposes. For example, while it is not surprising that a business ana lyst has to process millions of sales records to determine a valid marketing strategy, even casual users at home need to navigate thousands of movies to find the perfect entertainment for a night in, browse hundreds of social media updates daily to keep abreast of tl1eir circle of friends, or scan through thousands of product reviews to find the right toaster to buy. Regardless of application, the medium chosen to represent the data governs the ease with which a person can perform a specific task using the information. This means that successful designers should adapt the data presentation based on what the user needs to do.

The best medium for many tasks and types of data is a visual representation­ after all, a picture is supposedly worth a thousand words! For example, a building blueprin t, a geographic map, and a digital photograph are generally best presented as 2-D pictures on the computer screen and 11ot as a list of coor­ dinates, colors, and shapes. Similarly, while text is the optimal presentation to convey a single number (such as the cost of a product, the distance to the super­ market, or an approval percentage), a visual presentation such as a bar chart, a line graph, or a scatterplot is often a better choice when conveying multiple related points in a dataset, such as average reviews for mu ltiple products, stock values over time, or the relation between income and years of experience in a job. This idea of data-driven pictures is called visualization and is defined as the graphical representation of data to amplify cognition (Card, 2012; Ware, 2013). Visualization draws upon the massive bandwidth of our visua l system to essentia lly allovv people to "use vision to think" and dates as far back as Wil­ liam Playfair's line graphs and bar charts from 1786, Charles Minard's flow maps from 1869, Florence Nightingale's rose diagrams from 1857, and John Snow's cholera outbreak maps from 1854 (Fig. 16.6) (Tufte, 2001; Friendly, 2006).

In terms of Norman's gulfs of action, a macro-HCI theory describing the dif­ ference between a user's mental model and an interactive system's state (Chap­ ter 3), visualization minimizes the gu lf of eva luation because a well-designed graphica l representation is optimized for many perceptual tasks. Based on th is concept, Section 16.2 first presents the typical tasks that people tend to conduct using visual analysis methods. Section 16.3 then reviews typical data types and examples of common visualiza tion techniques designed for them. This example­ based framev.rork is necessary since the visualization discipline is young and still lacks specific macro-HCI theories for selecting the optimal visual representation

16.2 Tasks in Data Visualization 553

See also:

Chapter 7, Direct Manipu lat ion and lmmersive Env ironmen ts

Chapter 8, Fluid Navigation

Chapter 10, Devices

Chapter 11, Communication and Col laboration

Chapter 15, Information Search

that will minimize the gulf of evaluation for a dataset and task. Instead, visual­ ization design is often empirical in nature.

Compared to the static visualizations of old, computer-based visualization has the added benefit of being interactive, which opens up vast opportunities beyond the static representations printed on paper. Similar to the above discus­ sion, an effective interaction method for a visualization minimizes the gulf of execution-the difference between user intention and system actions-by enabling the user to easily carry out the task. However, interaction for visualiza­ tion differs irt many ways from typical interfaces and user applications.

Finally, much has happened in the more than two decades since the visual­ ization field was established at the end of the last century: Computers have become faster and evolved into new forms ranging from smartphones and tab­ lets to wall displays and tabletops, our society is awash in a deluge of data drawn from every discipline and domain, and a new generation of mobile and ubiquitous computing is turning our world into one where computing has dis­ appeared into the fabric of everyday life (Dourish and Bell, 2011). This means that many of the foundational principles that visualization researchers tradi­ tionally have held to be true no longer are. Section 16.4 reviews the challenges facing both researchers and practitioners in data visualization.

16.2 Tasks in Data Visualization

Why do people want to interact ,..,ith data? A pragmatic designer will start with the tasks that users want to perform in order to decide how to support those using interactive visual representations. Determining a standard set of such data analysis tasks has been an active area of research in the visualization community for the past two decades. One of the formative efforts i.t1 this venture was Shi1eiderman's visual infor,nation-seeking rnantra from 1996: "overview first, zoom and filter, then details on demand," which still accurately captures the high-level sensemaking process (Klein, 2006) that users engage m when mteracting with data. Amar et al. (2005)

554 Chapter 16 Data Visua lization

approached the problem from the other direction (i.e. from the bottom up instead of from the top down), deriving 10 low-level analytic tasks that people commonly perform: retrieve value, filter, compute derived value, find extremum , sort, deter­ mine range, characterize distribution, fu1d anomalies, cluster, and correlate. Mun­ zner (2014) since filled in the gap between high-level sensemaking and low-level analytic tasks using a typology of abstract visualization tasks, which focuses on the why, what, and how of engaging with data at all abstraction levels.

While these efforts lay the necessary theoretical foundation for how users engage with da ta, they do not provide concrete guidance for designers looking to build novel visualization tools. To achieve this, this section presents a taxon ­ omy of interactive dynamics that combine the analysis task with the practical operations that users need in their visualization tools (Heer and Shneiderman, 2012). The taxonomy consists of 12 task types grouped into three high-level categories, as shown in Box 16.1: (1) data and view specification (visualize, filter, sort, and derive); (2) view manipulation (select, navigate, coordinate, a1ld organize); and (3) process and provenance (record, anno tate, share, and guide). These thiee categories incorporate the critical tasks that enable iterative visual

BOX 16.1 Twelve task types fo r visua lization organized into three high-level catego ries (adapted from Heer and Shneiderman (2012)).

Task Categories

Data and view specification

View manipulation

Process and provenance

Task Types

Visualize data by choosing visual encodings

Filter out data to focus on releva nt items

Sort items to expose patterns

Derive va lues of mode ls from source data

Select items to highlig ht, fi lter, or man ipulate

Navigate to examine high-level patterns and low-leve l detail

Coordinate views for linked exp lorat ion

Organize mu ltiple windows and wor kspaces

Record analysis histories for revisitation, review, and sharing

Annotate patterns to doc ument findings

Share views and annota t ions to enab le collaboration

Guide users through ana lysis tasks or stories

16.2 Tasks in Data Visualization 555

analysis, including visualization creation, interactive querying, multi-view coordination, history, and collaboration.

For each of the 12 task types described here, examples are given that sho,v­ case the idea using real-world and predominantly commercial visualization tools. While this is by no means an exhaustive survey, these examples give prac­ tical and operational evidence of how designers can model their interfaces to support sensemaking of large-scale and complex data.

16.2. 1 Data and view specification Core functionality of any data visualization tool includes basic operations to visualize data using a visual representation, to filt er out unrelated information, and to sort information to expose patterns. Users also need to derive new data from the input data, such as normalized values, statistical summaries, and aggregates. These four task types can be explained as follows:

• Visualize data by choosing visual encodings. Not surprisingly, selecting a visual encoding for a particular dataset is the most fundamental operation for a visualization tool. A common approach in practical visualization tools is to simply provide a palette of available charts, allowing users to easily pick the chart most appropriate for their data (Fig. 16.1). Microsoft Excel and Tableau

••• O;iit;t. M• ~IC$ ' l'oOel – c..,_. 'fu t (bltl l Clnhlll S,,,.r1t AnM'l11 0.

-. NH .. CUtNumbu Mir ~tl,Mb(,tl

* CueMlm~? 0 (tou n1ry -""' * f 13 ,.. ,~, -""' • Htd° forMIIII. -"""' -"""""' .- Hill,,_. -.. , "' … "'_ … -·­,.. 1"¥1'111 –· . … • ()rlglnal 0,-W • Ye•

•·

;l

0 Wi'l'_.~,IMI 0 ~ ~lftll — ~olltttolds • MUSIi,_~

-Vc• (botl)

""'"' *'b

"'""' A – -·­·-

" …

·­… -·-

f'I On• souru I She~ l I tai tfl tD

…. ..

om

.. •

•

•

SS,_.,, l •-lit-11……,• suw."'t ……. u!Lt..-dll<•,"6

FIGURE 16.1

— – –

…. , .. ,.. ,.,. , … _ ,.,., ___ ,….

-__ ,..

■A-

tlli Sho. Mt ….. …_ -""'"I -.. ~ u._ – …. –

l~"il'f!'d JJ i="l hL IJt Q i@a!. i-;.,

■–■-, ll1i1 == • lillII . ~ . . . – · .. ·-1 ':: ":'♦- =-" · ~ -· • ~1AlmCOCrA ,. ,.,…,_. .,,.

D ~ , .,, • …….,. -~ o.-….. -~ o.:-·­·­·-■ … YO,-IIEJIIG,l,L ·­·-• …….. l'Cltruc.-&.&-. ….. fl "'"19M&a6' …… ~~~ • ………….. e.RtOl …… …,…,.. ·-…… . …….. D CMflflf,A,Ntu ■ CA-e.MOS D ~ Na'IC D ClnOll'OW~I ….. ·-• c- 11 """1l», II.,.,_

.. ' ' ~ . Visua lization palette (upper right) in the Tableau Desktop app lication for a dataset of shark attacks . The "show me" feature in the tool (Mackin lay et al., 2007) will automatica lly highlight the suitab le charts that can be used for the selected data.

556 Chapter 16 Data Visua lization

both provide such palettes; in addition, Tab leau also has a novel feature called "show me" that automatically selects the most appropriate visualization given the structure of the data (Mackinlay et al., 2007).

• Filter out data to focus on relevant iten1s. While the overview of a dataset is often important in orienting the user in a visualization, eliminating unrelated in­ formation from the view is critical as the user starts to investigate the da ta in detail. Several methods for filtering exist, such as direct ly lassoing important objects (Choi et al., 2015) or selecting intervals and values on data dimensions using dynamic queries. Fig. 15.10 shows a hotel search interface on the Kayak trave l website with an integra ted filter interface. The interface allows for dy­ namical ly querying the hotels that match filtering criteria by changing range sliders for price intervals and selecting features by checking boxes (review scores, free breakfast, free internet, etc.). The results update dynamically as the filters are changed.

• Sort items to expose patterns. Ordering da ta items according to some dimen­ sion, such as age, income, or price, is vital in exposing hidden patterns in the data. Sorting a list of items is often easily performed by clicking the header category; toggling reverses the order.

• Derive values of 1nodels fro1n source data. Original datasets can often be aug­ mented with data computed from the original, such as statistics (e.g., mean, median), transformatio11s, and even powerful data mitung metl1ods. In fact, calculating derived data as part of an interactive system with a user in the loop is a nascent but growing research area called visual analytics (Keim et al., 2008), where compu tational methods work in synergy with the user.

16.2.2 View manipulation Much of the value of visualiza tion comes from being able to manipulate the view on the screen, including tl1e ability to select items or regions, to navigate the viewport's position on a large visualization, to coordinate multip le views so that data can be seen from multiple perspectives, and to organize the resulting dashboards and workspaces.

• Select iterns to highlight, filter, or rnanipulate. Pointing to an item or region of interest is common in everyday commurucation because it indicates the subject of conversation and action. In a visuali zation tool, common forms of selection include clicks (by mouse or by touch), mouse hover, and region selections (e.g., rectangu lar and elliptical regions or free-form lassos) (Fig. 16.2).

• Navigate to examine high-level patterns and loiv-level detail. Visualizations often contain more information than can be comfortably shown on screen, either

~ D , Po,nte, Rect Ellipse

~ '-.. N t.asso Straight Polyt,ne

,Jllea ~M eajoo • /… a,.. Rotate ~i nt.Jt..byAre

rf/. ©.. AByData R~ LeJIS Circle n

~ Borckr'litle lnclusn·e

FIGURE 16.2

() Jondrette

16.2 Tasks in Data Visualization 557

(") Ftuily (") Child2 (") BJl!.orel

1:, Child1(")G. M~iuf

0 Combeferre

._ __ _ olras

GOToche deli"crs :',lams' farewdl i,ner for Cosette 10 Valjean ' – ,byl,nus—,– =———–l n1,@,ttf''(l!) MDe.Vaubois

Ttf.nM qu•siM (") FMOritt !l6cl), (") Gueulen Tholomyes (") F

(") MPe.Gjjloj)~d Epor,ine u ·•u ~.Bln"'f'lnd O . .

Listollet P<1nlmet') Ja•1trt O Boulatrlltlt

(") Mootpam~ . F~ n• t) Voljoa,, (") Mme.Thena r (") ze

(") Sc~ •""———— -.-..,,, r Dahlia M g'M,

0 EP.!)lline '1ops the robb;ng ofValjean's.Jleuse

~1 ~ ~,..:,: (") , -~~ l•u (") ~.sib,p liu

r Judgo (") Fau..0.00 1119""~• 0 Motherl,m~c,nt O Mil 8 ~ · ,·

0 I ..__ e, a,….1s 1ne

sa~ ~ • 0 enitdieu I Gervais Myriel (") Brf'lt t

(") Cochepaillo

0 Champ:ercier 1

Valjcan's conscience Gritlie

(") OldMan

(") C~ tn dtlo (") Geborand ~ J C c,Q 1i,tapoleon

0 PttpttUt

Select ion tools and data-aware annotat io ns in an interactive node- link d iagram representation of the social network for all of the characters in Victor Hugo's Les Miserables. Characters are linked together if they appear in the same chapter in the book. The textual annotations are connected to nodes using red lines and stay connected as the graph layout changes. The toolbar on the upper left is part of the VisDock toolkit and provides tools for annotation, navigation, and selection (Choi et al., 2015).

due to the sheer 11umber of pixels or just due to high visual clu tter . For such dense information spaces, navigation operations such as pan and zoom allow the user to control the size and position of the viewport on the visualization (see the navigation tools in the tool dock in Fig. 16.2 or the map contro ls in Fig. 15.11). Not surpri singly, zooming and panning operations have now be­ come common in many conventional user applications, such as Google Maps, Adobe Photoshop, and Microsoft Word.

• Coordinate vie-ivs for linked exploration. Since each visualization technique has its own strengths and weaknesses, practical visualization tools often include multiple views of the same dataset so that each view illustrates a specific as­ pect of the data. When using multiple views in this way, such as in a visu­ alization dashboard (Few, 2013), it is customary to coordinate the views so

558 Chapter 16 Data Visualization

284 Data Breaches Industry

GOYem~ & Mih'lary 32 ~leol~rt& Mc(liQDI Provide 18 Fina!'ICial and lns,,iranoe Serv- 11

t.,.. e o….,,_ Aug 20. 201-1 CIIC011munity Health Systems May 22. 201◄ ClMcntana Department of Public Health and

Human Services Othf!f 11, : May 21. 2014 $ eBay

Ed~IOl'IOl ln$'lit1,11ions 21- i

,., "' '" .. ,,, .. ..

AetaU &Merchant 6- Apt 28. 20l◄ $ AQ t1-1xw,,,. 1 More • 20 -<0 60 AflflS.2014 ® Bu'.garia Citizens

# of Records … Feb 13, 2014 SI F0 1bes • Jon 27,2014 8 Mic:haer, Stores. Aaron Brothers Jan2s.201• ® Michaels Stores Inc. Jan 01, 201• 8 Sk!pe

– Dec 2S. 2013 S Snapchat ~ 41 -n-' -;–,- Dec 06. 2013 8l Ho-izon Blue Cross Blue Shield of New JersQy

, I )

JOl.: 1~ .. ~ 2QM ~ J" Dec 06.2013 Ci Horizon Healthcare Services, Inc. (Hortzon Blue Source Cross Blue Shield)

1.1:alic10UsOutsider ~s: = =~ Noy '27. 2013 CIMericopa county Commun ity College District PhySieall.0$$ 50 Nov2S.2013 CS,Evernote

AecicbWalD1;11a 1.0$$ 21- Nov 13, 2013 8 MacRumors fOfurns

Ma!JciOls Insider tia Nov 13. 2013 8 m1.scogee county lklk:ncwm 11 Novo<1.,201aerarget

suneS()OnSOred o 1 i:i.,ws • 50 ,00 NoYO<l.2013 e co rpo rateCarOnline.com

00 00

Date

., ,. " " ,. ' • .,,. … … J:in

FIGURE 16.3

OIIDOX

Tn>e l<!Mtitylll~ 1s_ llllil_

Aoc:ountAoem 3- flrenc:.el Aoeess 1-

Eti!.leMial D3ta 1 ■ Nuis:n:ie 2■ 1

'5~ I 10 21) 30

Data Source Prlvscy RlghtsClearinghrus,e 102 _ _ _

8reat:h lEM!I Index 22 • W 100 1:,0 111)

Rocord Typos

Nome o Pt-rsonat, ldcmifying lnf<.lfm_ 14

Social $e::Ur1ty NIJl'l'lbef 9

DeyofBlllh 8 emeH s

Adctess ri

Password 4

Credit Card NLfflber 2

Log 2- ! Medlcel Aecord, soc •

FIN 2 8 Rr,,,,,11 i ,.,…,. • 1 0 ,0

…. .r-.n …. '

Exploring 284 data breaches in the United States using Keshif (ht tp://keshi f .me/), a multi-view visualization tool that shows different aspects of the data in separate views (Valcin, 2016). Selecting items in one view highlights them in others; for examp le, the user is current ly hovering over the bar for "70k-300k" in the view tit led "# of Records," which causes those 124 breaches to also be highl ighted in orange in other views, includ ing in the timel ine at the bottom.

that selecting items in one view high lights the item (or related items) in other views (Fig. 16.3). Also see Section 12.3.1 for more on this.

• Organize rnultiple windoivs and ivorkspaces. While involving multiple views of a dataset allows users to explore complex data using straightforward and familiar visuali zations, this also introduc es the need for users to organ ize and lay out the views to fit their needs. Man y tools allow for dragging and drop ­ ping views to achieve this, such as the Keshif tool shown in Fig. 16.3.

16.2.3 Process and provenance If the previous two categories of tasks deal with the mechanics of creating, manipulating, and view ing vis uali za tions, the third category encompasses higher-level tasks for scaffolding, interpr eting, and documenting the

16.2 Tasks in Data Visualization 559

exploration process. More specifically, the tasks here involve the ability to record the analysis, to annotate regions of interest in a visualization, to share views with colleague s, and to guide other s through presentations of the analysis outcome.

• Record analysis histories for revisitation, review, and sharing. Visualization tools do not only help user s collect insights from their data, they should ideally also support mecharlisms to record these insight s as well as the path lead ­ ing up to them. One approach that several tools provide is an automatically recorded history of interactions, allowing the user to review and revisit the exploration and even shar e it with others (Fig. 16.4).

• Annotate patterns to docun1ent findings. Most visuali zation s use data in a read­ only fashion since the goal is to let the data inform the user 's exploration, but some tools allow for adding metadata in the form of textual or graphical annotations associated with the visualization (Fig. 16.2). Textual annotations constitute labels, captions, or comments, whereas graphical annotations are sketches, highlights, or handwritte11 notes. To be truly useful, annotations should be data-aware so that they are associated with underlying data poiI1ts and not just drawn as a transparent layer on top of the visualization (Heer and Shneiderman , 2012; Choi et al., 2015). Drawing annotations on such a tran sparent layer make them meaningless when the visualization is filtered or reorgaruzed.

• Share views and annotations to enable collaboration. Analyzing data is often a social activity involving multiple users working together (Heer et al., 2009), either in de facto teams or in loose constellations of people on the

x Worksheet History

Filters:

Graolw: T'll)e:

8ool<matks: 0 Show Only

"40r"-.. , s..ltw c..rt, ,I 1.4").P't l,w.J,OtS

ttill l,.:t,.'6-J .._._.~I ~ m,w N.O.JN -. 1.1n,15,4 z.,n,•10

Add Inventory

( Expo,t ] I Reset ] !<] g

F GURE 16.4

– II~ 11.1

ShowMe ! Move Market Size to Add Product to Coums Coums

Graphical history interface using thumbnails of previous visualization states organized in a comic-strip layout (Heer et al., 2008). The labe ls describe the actions performed.

560 Chapter 16 Data Visualization

U.S. Shark Incidents by State

-·­·-

-.. .,

•

•

•

• •

,. • :l

• •

• •

–

—–

—–~

< • > V

+

•:rr.o….. …

• ——- •er-:-.. ,…

Y-• Ollle • $Ille l•- rllal Type "9,.. ~. A~"1 :tu u,..- 1"1otw "'"""'°"".,._c.co, DNca.&. Urco-hd"""""' ' 10 , ,.._.,.,, 20H :e,…,_ l…w.. UN,._…11-•~ u,p,ocuc….,. , l' w. ~ 2014 2014,.Qt.ft S…. C•-. ,_, ha, C~ C- IJ "'1'fo.bdll'lllll• 1 10 M 9M,oi9 ~

FIGURE 16.5

0

,._,_ r.•r, ~ l X — ___ .,. •- I I –Nfl n

w _ –• ,.,.. .. Martll_, ~

·-· Cllliro.N ■

$to,il'IC.l'OIIN I MorlhC.folh I

, .. ., I , … J … ., I

NNYO!lt I O,t9,;n I ,_, ""~In• I

PuffloRiCOI

•ti11•• I …… 1

I 0

_,._ _,,.,_ • —- • 5No. lntl0tl9 A … –· _ ….. c …. • t•on•f.,.11 . , … ,

500 ,.,. –0

-~–~-""' WNf._lla!IIO lt ll ..__ . ._….,.,

,., … ~~~· Q

I t,lr,l) 2hd.l~ -… .. C-• .,_ .. —· , ….

Y- IU.. 7014

o— o …. m: ••01 au,,.. a o— o , ….. ..,.,., …. ~•–likfJ~ ….. b!. -….. … i.c-,, ,

I! au.,ne,, … M( …

;i: U,c,rorr,ct;IClll,;llktll ·-{E fa. ·" n,.. ………. IZ • IZ ' :z ,r…e,, Agt1of1doa . ., o— o

Spotfire visualization dashboard of shark attacks published on the web. Users can interact with the dashboard, causing views to update dynamically. The tool also allows for application bookmarking (storing the state for specific insights) as well as sharing the analysis on social media platforms such as Facebook, Twitter, and Linkedln.

internet. The implication is clear: to support the ana lysis life cycle fully, visual analytics tools should support social interaction. This could include simple functionality to export shareable formats of charts (PDF, PNG, JPG, etc.) and datasets (CSV, JSON, XLS, etc.) from a visualization tool as well as more advanced sharing mechanisms such as application bookmark­ ing and publishing visualiza tions on the web (Fig. 16.5).

• Guide users through analysis tasks or stories. As visua lization tools become increasingly available to casual users looking to get insight into their own data-such as their social networks, personal finances, or local communities­ there is also an increased need to guide these novices through appropriate approaches to analy ze their data. Simjlarly, carefully crafted data stories can help explain even complex phenomena using a combination of visua lization s, annotations, and textual descriptions (Fig. 16.6).

London's 185-1 Cholera Outbreak: Data ~1apping Halts an Epidemic

( An Ou!bfeak Begins Colecting the Data

Armed with this information, he "·ent to city authorities.

• C,.•,,n

• l.kfoulnl

I No.~AM … I

lo , I ::

FIGURE 16.6

Mapping the Results

16.3 Visualization by Data Type 561

Snow's Analysis Focus on Broad St.

Encing an Epidemic

>

Web-based visua lization of London's 1854 cholera outbreak showing physician John Snow's use of visualization to find its source. This visualization was created in Tableau using its Story Points feature, which allows users to build a narrative from data. The horizontal list of five boxes at the top of the display are the main points in t he story, and v iewers can be automatically guided through the story by moving to each poin t from left to right.

16.3 Visualization by Data Type

The visualization field currently lacks a unified theory that can recommend the optimal visualization technique given the data type to represent and the tasks that the user wants to perform. Furthermore, most data typ es

Collepals.com Plagiarism Free Papers

Are you looking for custom essay writing service or even dissertation writing services? Just request for our write my paper service, and we'll match you with the best essay writer in your subject! With an exceptional team of professional academic experts in a wide range of subjects, we can guarantee you an unrivaled quality of custom-written papers.

Get ZERO PLAGIARISM, HUMAN WRITTEN ESSAYS

Why Hire Collepals.com writers to do your paper?

Quality- We are experienced and have access to ample research materials.

We write plagiarism Free Content

Confidential- We never share or sell your personal information to third parties.

Support-Chat with us today! We are always waiting to answer all your questions.