1. A TALK THAT SHOWS SOME BAAAD CHARTS,
SAYS WHY THESE ARE BAAAD CHARTS, AND
INCLUDES DISCUSSION OF HOW
THE CHARTS COULD BE IMPROVED
Daniel Brooks Wright
Research Department
Aspire University, Redwood City, CA, 94062, USA
Lots of other information that you will probably
not write down, is too small to read, is in a color
that is tricky to see against this un-necessary
background, and may be mes-spilt.

2. A TALK THAT SHOWS SOME BAAAD CHARTS,
SAYS WHY THESE ARE BAAAD CHARTS, AND
INCLUDES DISCUSSION OF HOW
THE CHARTS COULD BE IMPROVED
Daniel Brooks Wright
Research Department
Aspire University, Redwood City, CA, 94062, USA
Lots of other information that you will probably
not write down, is too small to read, is in a color
that is tricky to see against this un-necessary
background, and may be mes-spilt.

3. Making Bad Charts Daniel Brooks Wright Research Department
Aspire University, Redwood City, CA, 94062, USA
Lots of other information that you will probably
not write down, is too small to read, is in a color
that is tricky to see against this un-necessary
background, and may be mes-spilt.

4. Making Bad Charts Dan Wright danielw@aspire.university

5. Confusing People with Numbers
“Concentrate on methods of data display that leave the viewers as uninformed as they were before seeing the display or, worse, those that induce confusion”
(Wainer, 1984, p. 137)
And how to do a bad PowerPoint.

6. Three Modes of (Mis)Communication
Numbers
Words
Pictures (Charts)
I will tell you how to present information badly (and then show how to avoid this!).

7. Number Miscommunication
Round inappropriately
Avoid guidelines.
Testing organizations publish this 
Do not understand the statistics
that you are presenting to others.
Avoid ways for making clear tables
(there are books for making good tables).

8. datatab <- matrix(c(11843,11201,11053,10945,11317,11546,11673,11294,11692,
35305,35395,35135,34999,35300,36163,35976,36017,36452,36078,
11669,11535,11537,11408,11712,11892,11521,11073,10372,10115,
35779,35857,35944,36398,36702,36857,36097,34487,34293,
25526,36903,28696,29681,29994,30321,30477,31209,32308,32586,
32094,32420,32977,33625,34143,34612,35040,33888,33747,
37395,38570,39293,40225,41649,43182,43951,43531,43832),
27273,28115,29257,30412,31459,32488,33730,34905,36020,36514,
ncol=4)
percap <- datatab/rowSums(datatab)
years <- 1992:2010
lines(years,percap[,1],lwd=2.5,col="grey10")
plot(years,percap[,1],xlab="Year",ylab="Proportion",las=1,col="white",ylim=c(0,.5))
lines(years,percap[,2],lwd=2.5,col="grey20")
lines(years,percap[,3],lwd=2.5,col="grey40")
lines(years,percap[,4],lwd=2.5,col="grey70")
text(2005,.365,"Bachelor's or higher")
text(1997,.08,"Less than HS diploma",cex=1)
text(2001,.25,"Some college")
text(1996,.38,"HS grad,")
text(1996,.35,"no college")
lagtab <- percap*NA
for (i in 1:4)
lagtab[,i] <- c(NA,diff(percap[,i]))*100
plot(years[-1],percap[-1,1],xlab="Year",ylab="Percentage change",las=1,col="white",ylim=c(-6,8))
lines(years[-1], lagtab[-1,2],lwd=2.5,col="grey20")
lines(years[-1],lagtab[-1,1],lwd=2.5,col="grey10")
lines(years[-1], lagtab[-1,3],lwd=2.5,col="grey40")
lines(years[-1], lagtab[-1,4],lwd=2.5,col="grey70")
text(1996.6,3,"Bachelor's or higher")
text(1994.8,6,"Some college")
abline(h=0,lty=2)
plot(years[c(-1,-2,-3)],percap[c(-1,-2,-3),1],xlab="Year",ylab="Percentage change",las=1,col="white",ylim=c(-1,1))
lines(years[c(-1,-2,-3)],lagtab[c(-1,-2,-3),1],lwd=2.5,col="grey10")
lines(years[c(-1,-2,-3)], lagtab[c(-1,-2,-3),2],lwd=2.5,col="grey20")
lines(years[c(-1,-2,-3)], lagtab[c(-1,-2,-3),3],lwd=2.5,col="grey40")
lines(years[c(-1,-2,-3)], lagtab[c(-1,-2,-3),4],lwd=2.5,col="grey70")
text(1999,.83,"Bachelor's or higher")
text(2002.4,.17,"Some college")
text(2001,-.78,"Less than HS diploma",cex=1)
text(2008,-.67,"HS grad,")
arrows(2001,-.68,1999.5,-.25,.1)
text(2008,-.79,"no college")
text(2004,.89,"Bachelor's or higher")
text(1999,-.6,"Less than HS diploma",cex=1)

13. datatab <- matrix(c(11843,11201,11053,10945,11317,11546,11673,11294,11692,
35305,35395,35135,34999,35300,36163,35976,36017,36452,36078,
11669,11535,11537,11408,11712,11892,11521,11073,10372,10115,
35779,35857,35944,36398,36702,36857,36097,34487,34293,
25526,36903,28696,29681,29994,30321,30477,31209,32308,32586,
32094,32420,32977,33625,34143,34612,35040,33888,33747,
37395,38570,39293,40225,41649,43182,43951,43531,43832),
27273,28115,29257,30412,31459,32488,33730,34905,36020,36514,
ncol=4)
percap <- datatab/rowSums(datatab)
years <- 1992:2010
lines(years,percap[,1],lwd=2.5,col="grey10")
plot(years,percap[,1],xlab="Year",ylab="Proportion",las=1,col="white",ylim=c(0,.5))
lines(years,percap[,2],lwd=2.5,col="grey20")
lines(years,percap[,3],lwd=2.5,col="grey40")
lines(years,percap[,4],lwd=2.5,col="grey70")
text(2005,.365,"Bachelor's or higher")
text(1997,.08,"Less than HS diploma",cex=1)
text(2001,.25,"Some college")
text(1996,.38,"HS grad,")
text(1996,.35,"no college")
lagtab <- percap*NA
for (i in 1:4)
lagtab[,i] <- c(NA,diff(percap[,i]))*100
plot(years[-1],percap[-1,1],xlab="Year",ylab="Percentage change",las=1,col="white",ylim=c(-6,8))
lines(years[-1], lagtab[-1,2],lwd=2.5,col="grey20")
lines(years[-1],lagtab[-1,1],lwd=2.5,col="grey10")
lines(years[-1], lagtab[-1,3],lwd=2.5,col="grey40")
lines(years[-1], lagtab[-1,4],lwd=2.5,col="grey70")
text(1996.6,3,"Bachelor's or higher")
text(1994.8,6,"Some college")
abline(h=0,lty=2)
plot(years[c(-1,-2,-3)],percap[c(-1,-2,-3),1],xlab="Year",ylab="Percentage change",las=1,col="white",ylim=c(-1,1))
lines(years[c(-1,-2,-3)],lagtab[c(-1,-2,-3),1],lwd=2.5,col="grey10")
lines(years[c(-1,-2,-3)], lagtab[c(-1,-2,-3),2],lwd=2.5,col="grey20")
lines(years[c(-1,-2,-3)], lagtab[c(-1,-2,-3),3],lwd=2.5,col="grey40")
lines(years[c(-1,-2,-3)], lagtab[c(-1,-2,-3),4],lwd=2.5,col="grey70")
text(1999,.83,"Bachelor's or higher")
text(2002.4,.17,"Some college")
text(2001,-.78,"Less than HS diploma",cex=1)
text(2008,-.67,"HS grad,")
arrows(2001,-.68,1999.5,-.25,.1)
text(2008,-.79,"no college")
text(2004,.89,"Bachelor's or higher")
text(1999,-.6,"Less than HS diploma",cex=1)

14. Word Miscommunication
Use flowery "thesaurus-style" writing.
Show your novelist (or artist) aspirations.
Don't worry about accuracy. Be overly confident and use causal phrases inappropriately.
Write small for slides and use multiple weird fonts.
Rules grammar ignored can be (spelleng two).
Write as if you are writing for someone with a PhD in statistics. Make it sound complex! (lots of colors)
Don’t bother explaining odd procedures.
Sound bored (use these effects to prove it).

15. From Wired (2003) (yes, I know that I am using PowerPoint)
Edward Tufte
From Wired (2003) (yes, I know that I am using PowerPoint)

16. Rules for Graphical Presentation
Wainer (1984)
1. Show the data
2. Show the data accurately
3. Show the data clearly
Tufte (2001, p. 105)
1. Above all else show the data
2. Maximize the data-to-ink ratio
3. Erase non-data-ink (chartjunk)
4. Erase redundant data-ink
5. Revise and edit

18. Knowledge of How Vision Works can be Useful
cols <- c("red3","chocolate4","green2")
pie(c(1,1.5,2),col=cols,lty=0,labels=c("Failing", "Almost","Proficient"),cex=2)

19. (Cleveland and McGill, 1984)

21. Insensitivity to red light.
Most common form of color blindness.

22. Handouts should work if photocopied without color.
Filters and Palettes designed for color-blind and grey-scales.

23. Failure 23% Almost 33% Proficient 44% 100%
Tufte's data-to-ink ratio

24. perc <- rev( c(4,6,13,18,23,24,11))
names <- (c("Teaching/lecturing","Psychologist","Marketing, sales, retail",
"Adminstration/management","Social, welfare, health",
"Clerical, secretarial","Other"))
par(mar=c(5, 14, 4, 2))
barplot(perc,names=NULL,horiz=TRUE,xlab="Percentage",xlim=c(0,25))
title("Where do Psychology Graduates go? ")
mtext(names,2,.5,at=seq(7.9,.7,length=7),las=1)

27. (from Wright & London, 2009, statistics textbook)
... a false third dimension. This is at best a waste of ink and is likely to make the graph more difficult to read. ... Wallgren and colleagues (1996: 71) suggest this is included because it 'symbolize[s] the triumph of data technology' over thought. When we politely criticize people for adding a false dimension to a graph, they often tell us it looks 'pretty'. 'Pretty' does not count. Information, clarity, and accuracy count. 'Pretty' is fine providing these other three are fulfilled.
(from Wright & London, 2009, statistics textbook)

28. According to SYSTAT manual (Wilkinson, 2000, p
According to SYSTAT manual (Wilkinson, 2000, p. 13) these false 3d pie charts “incorporate nearly every visual illusion discussed in this chapter.”

31. Here is what is on prospects.ac.uk

32. from Edward Tufte (2001)
“may well be the worst
graphic ever”

34. x <- c(28,29.2,32.8,33.8,33)
years <- 1972:1976
dotchart(x,years,xlab="Percentage of Students over 25",xlim=c(0,50))
plot(x,years,xlab="Percentage of Students over 25",xlim=c(0,50),las=1,ylab="",ylim=c(1971.3,1976.7),bty="l",pch=19)
for (k in 1:5) lines(c(-1.70,x[k]),rep(years[k],2),lty=2)

36. The “aaaah” factor???

37. You: “Dan, this so-called Chartjunk that you do not like IS sometimes useful. It looks pretty and draws attention to the reader.”
Dan: “It interferes with communication of the information. It belittles your audience to say that they will only look at your graph if it is pretty. If the data are ugly, the data are ugly.”
If you desire, show your clear, informative graph and then show …

39. Another Example
What is Dan’s favorite baseball team?

42. While my shirt shows that I am a fashionable trendy person with stunning dress-sense, it does not communicate information about “which baseball team” well.

43. Making a Good Chart
Same rules for numeric, verbal, and graphical modes of communication.
Clarity
Accuracy
Don’t add extra junk
Pretty doesn’t count, except in fashion!

44. The End