214 Reporting quantitative data - UC Berkeley School of...
Transcript of 214 Reporting quantitative data - UC Berkeley School of...
Needs and Usability Assessment 11.04.08
I214 Repor ting quantitative data
November 4, 2008
(with thanks to Gary Kass and Edward Tufte)
Needs and Usability Assessment 11.04.08
Why graphs?
Ex: The Harris Poll #100, Oct 15, 2007
American adults have their choice of many differentcuisines … From pasta to burritos, the choices are almostendless. But what do Americans choose when eating out?American food, of course!
Over one-quarter (28%) of U.S. adults say A merican food is what theyare most l ikely to choose i f they had the choice to go out to arestaurant and eat one type of food. Just under one-quarter (22%) saythat they would most l ikely choose Ital ian while 17 percent wouldchoose Mexican while 16 percent would choose Chinese i f they had thechoice to go out to a restaurant. Japanese is the next choice as sevenpercent say they would choose this type of cu isine. Much furtherdown the l ist are Indian (2% say they would choose) and then Frenchand Middle Eastern cu isine (1% would choose each). Final ly , fourpercent say they would choose another type of food.
or…
Needs and Usability Assessment 11.04.08
Needs and Usability Assessment 11.04.08
Reporting results the easy way
SE9. How confident do you feel about your own searching abilitieswhen using a search engine to find information online?
Based on Internet users who have a used a search engine[N=1,165]% 48 Very confident 44 Somewhat confident 6 Not too confident 2 Not confident at all * Don’t know/Refused
http://www.pewinternet.org/pdfs/PIP_Search_Questions.pdf
exact questions and answers
Needs and Usability Assessment 11.04.08
Tabular displays
n= 12n = 10n = 2No ans.
100n = 308
100n = 190
100n = 118Total
69N=127
87N=165
40N=47Dissatisfied
31N=96
13N=25
60N=71Satisfied
Total%n=320
Grads%n=200
Undergrads%n=120
Comparing dependent to independent variables
Needs and Usability Assessment 11.04.08
Good information graphics
Efficientminimizing non-data ink
Meaningfuldisplaying informationrelevant to theargument
UnambiguousSelf-explanatory, or atleast difficult tomisread
http://www.prb.org/Artic les/2007/623Urbanization.aspx
Needs and Usability Assessment 11.04.08
5 principlesfrom E. Tufte
Visual Display of Quantitative Information,p 103 and 105
Above all else show the data
Maximize the data-ink ratio
Erase non-data-ink, within reason
Erase redundant data-ink
Revise and edit
Needs and Usability Assessment 11.04.08
Components of a chart
labelingTitleaxis titles and labelslegendsnotes
scales of X, Y axis
graphicrepresentationsof data
Needs and Usability Assessment 11.04.08
Area chart
How the dependentrelationship of parts ofa whole varies over anindependent variable
Percentage: sum of the twoalways represents 100%
Cumulative: sum variesaccording to elements
Needs and Usability Assessment 11.04.08
Column or bar
Change within smalldata sets that vary overa nominal (ie, category)or interval scale
Column: vertical rectangles
Bar: horizontal
Needs and Usability Assessment 11.04.08
SegmentedcolumnA columnchart thatrepresents thepart-wholerelationshipsof elements
http://www.stanford.edu/group/siqss/research/time_study_files/ProjectReport2005.pdf
Needs and Usability Assessment 11.04.08
Frequency polygon; Histogram
Overall shaperepresentsdistribution ofelements
Needs and Usability Assessment 11.04.08
Line charts
Used to interpolate betweendata points and extrapolatefuture valuesTrends or change over timeLooking for correlationsbetween variables
Lines: Connect "real" data points
Curves: represent functionalrelations between data points orto interpolate data
X must have numerical values!
http://www.fivethirtyeight.com/2008/11/google-traffic-suggests-mccain-not.html
Needs and Usability Assessment 11.04.08
Pie chart
Good for displayingapproximate percentagesat a single point in time
Caution:Makes exact comparisonsdifficult
You cannot necessarilycompare pie charts to each other
Maybe use a bar chart instead?
Needs and Usability Assessment 11.04.08
Scatterplot
One dimensionalShow measurements over time
Two-dimensionalConvey an overall impression ofthe relation between two variables
Some caution:Not good for following trends
Should not be used with morethan one independent variable
Needs and Usability Assessment 11.04.08
Analysis
Looking for significantdifferences acrossimportant groups
Looking for correlation– inferring causality
Polls includingmobile phones
Needs and Usability Assessment 11.04.08
Maps
Placingdata onphysicalspace
Needs and Usability Assessment 11.04.08
Representing change withsmall multiples
Smallmultiples
“Graphicaldepictions of
variableinformation
that sharecontext, but not
content.”Edward Tufte
Tufte, E. Envisioning Information. Cheshire, CT: Graphics Press (1990).
graphic: www.uxmatters.com/MT/archives/000044.php
Needs and Usability Assessment 11.04.08
Reportingguidelines
The readershould be ableto reconstructwhat you did
Show thequestion
Include all dataie, percentagesand numbers ofrespondents
n= 12n = 10n = 2No ans.
100n = 308
100n = 190
100n = 118Total
69N=127
87N=165
40N=47Dissatisfied
31N=96
13N=25
60N=71Satisfied
Total%n=320
Grads%n=200
Undergrads%n=120
Table 1: Opinion on foodAre you satisfied or disssatisfied with the bagels at the FreeSpeech Movement Café?
Needs and Usability Assessment 11.04.08
Reporting: last words
Graphics are visualmetaphors.They are rhetoric.
Do you know the storyyou want to tell?
…the story youraudience is seeing?
Needs and Usability Assessment 11.04.08
Appendix A:Good places to look for examplereportsPew Internethttp://www.pewinternet.org/pdfs/PIP_Searchengine_users.pdf pp. i-
ivhttp://www.pewinternet.org/pdfs/PIP_Future_of_Internet_2006.pdfhttp://www.pewinternet.org/pdfs/PIP_Search_Questions.pdf"What Do Americans Do on the Internet?“ Appendix B
Harris Interactive Pollhttp://www.harrisinteractive.com/harris_poll/
Stanford Time Studyhttp://www.stanford.edu/group/siqss/SIQSS_Time_Study_04.pdf