Admin G/B/U Few R

Lecture 4:What Graphs Do and Making Bar Charts

February 12, 2018

Admin G/B/U Few R

Overview

Course Administration

Good, Bad and Ugly

Few, Chapters 6

Bar Charts in R

Admin G/B/U Few R

Course Administration

1. Will return proposal comments during programming

2. Rosa has graded problem sets – thank you

3. Grades posted?

4. Missing anything from me?

Admin G/B/U Few R

Next Week’s Good Bad and Ugly

Monday by 9 am. Earlier is ok.

• Adam Brooks

• Gulfishan Khadim

Admin G/B/U Few R

This Week’s Good Bad and Ugly

• Kelsey Wilson

• Nathan Rupp

• Haley Dunn

Admin G/B/U Few R

Kelsey’s Example

Admin G/B/U Few R

Nathan’s Example

Chapter 4 | The Efficient A

SEAN

Scenario

93

4

Figure 4.1 ⊳ Trends in energy intensity and GDP per capita in selected countries, 1980 - 2011

Indonesia

Malaysia

Philippines

Thailand

Vietnam China India

toe per 1 000 dollars (2012, MER)

Dollars per capita (2012, MER)

10 000 8 000 6 000 4 000 2 000

1.6

1.4

1.2

1.0

0.8

0.6

0.4

0.2

OECD Singapore

Japan

Malaysia

10 000 20 000 30 000 40 000 50 000 60 000

0.1

0.2

0.3 1980

1990

2000

2011

Note: GDP is measured at market exchange rates (MER) in year-2012 dollars.

0 0

0 0

© OECD/IEA, 2013

Admin G/B/U Few R

Haley’s Example

Admin G/B/U Few R

Few:Fundamental Variations of Graphs

Admin G/B/U Few R

Today

1. Types of graphs

2. What you can communicate, by graph

Admin G/B/U Few R

1. Types of Graphs

• Points

• Lines

• Bars

• Boxes

• Shapes with varying 2-D areas

• Lines

Admin G/B/U Few R

Why to Avoid 2-D Sizes

Admin G/B/U Few R

Graph Design Solutions

As we go through these, we’ll discuss policy examples

1. Nominal Comparisons

2. Time Series Designs

3. Ranking Designs

4. Part-to-Whole Designs

5. Deviation Designs

6. Distribution Designs

7. Correlation Designs

8. Geospatial Designs

Admin G/B/U Few R

1. Nominal Comparisons

• Use bars: horizontal or vertical

• Or points to compare values

• Possible for not “too many” values

Admin G/B/U Few R

For Example

Admin G/B/U Few R

2. Time Series Designs

• Present data over time: months, days, hours, years, decades,...

• Almost required to use horizontal axis left to right for time

• And usually a connected line, with or without dots

• If time intervals are not consistent, then maybe dots or bars

• Lines indicate connection between observations, so watch outif you’re using them in another context

Admin G/B/U Few R

For Example

��������� ������ ��������

������������ �� ������������������������� ���!"#�$%�&'�%�&'���������������������()*'&+��'�((�'+�+&���,���*-��������*��%./� ���

012345678997999:;<3=4<>9?@6<;A2BC<D

EF<A=GFH;:I22J=;KFBGBA;LMGN22KFBGBA;LOPQO2B;R

O2:;:;<3=4<S:2HT1NJGH;<?A2HU

OG:T;4<3HHG:RVWXY2H;QMPZ[WXYLK;1\78[96]OZP̂6\?86_PE9?99>9?99̀ Dab;B LYFJN Lc2d L OT4AGb 6?e8X]fZ:G4F2 5@?9eEFgRF;=C 9?h8̀KFBGBAFG=B;d<7A2HbG:F<2B<GBCH2:;�))� �))' ���� ���' ���� ���'����������(�(�i�i�������WXYIGH;<T1NJGH;<?A2Hf]=GR2g;:5@999GHGjFBJS:;;JGH;<2BklmIGH;<?6]=GR;:]=GR6b=GR;:JGH;<2BWXYIGH;<7d;NGg;4N;1;<4??? P3b;:OG:F2@8]=GRP3b;:OG:F2@8JGH;2B=FB;G0A4F2BIGH;<IGH;<2BWXY???P3b;:PHG<NK=G<N6]=GRP3b;:PHG<NK=G<N6JGH;2B=FB;GK:;;IGH;<2BWXY??? ]2T;H2B]=GRb2T;H2BJGH;<2B=FB;?I22B;bFAGCg;B43:;424:GFBGBC???KFJN4FBJKFJN4421;A2H;4N;3=4FHG4;dG::F2:FBHGBR2S23:nJN4FBJ??? 0BFH;XG44=;6?6]=GR0BFH;XG44=;6?6JGH;2B=FB;G0A4F2BIGH;<IGH;<2BWXY???WXY[P3HHG:RS2:WXY2H;LMGN22KFBGBA;N44b<[ffnBGBA;?RGN22?A2Hfo324;fWXYfeCGR<GJ2LpF;d4N;1G<FAklm<42ATANG:42BMGN22KFBGBA;?qNGBJ;4N;CG4;:GBJ;7ANG:44Rb;GBCA2HbG:;WXY2H;GJGFB<424N;:A2HbGBF;<?WXYP42AT]:FA;LWXY2H;P42ATr324;>_?P?[QMPZDLOG:T;4sG4ANN44b<[ffddd?HG:T;4dG4AN?A2HfFBg;<4FBJf<42ATfT1NWXY2H;<42ATb:FA;4G:J;4:GF<;C42t89S:2Hth8G4uXqqGbF4G=????WXY2H;1;G4<;vb;A4G4F2B<7;vb;A4<w<2=FCC;HGBCwS2:N23<FBJFB695x????WXY2H;<42ATSGF:gG=3;;<4FHG4;:GF<;C42th9S:2Ht6@G4OWO]G:4B;:<?WXYWXY2H;P42ATr324;LKFBgFj?A2HN44b<[ffnBgFj?A2Hfo324;?G<Nvy4zWXY{GBL6eL5x98[59]O7WXY2H;E;A=G:;<KF:<4r3G:4;:695xEFgFC;BCX3<FB;<<sF:;?{GBL68L5x9x[990O7WXY2H;ZvbGBC<Y;G=4NRY2H;K;G43:;<|}B4:2C3A;<Q;dWdFT<;4~]:2C3A4<X3<FB;<<sF:;?{GBL5\L5x59[e@0O7WXY2H;[klmL_P[ZG:BFBJ<0BG=R<F<[r87695�XR4N;Q3H1;:<[{GB3G:R5\7695xqGbF4G=q31;?WXY6\?86�9?\h>WXY2H;DLO2:BFBJ<4G:ddd?H2:BFBJ<4G:?A2Hf<42AT<fvBR<fT1Nfo324;?N4H=O2:BFBJ<4G::;G=L4FH;<42ATo324;<7J:GbN<7GBCFBC;b;BC;B4GBG=R<F<S2:klmT;;bR23FBS2:H;C?P4GR3b42CG4;dF4NWXY2H;<42ATb:FA;GBCGBG=R<4:G4FBJ<2BO2:BFBJ<4G:?A2H?

q2HbGBRWXY2H;F<GN2H;13F=CFBJA2HbGBR1G<;CFB4FB5\e�G<WG3SHGB�X:2GCFBE;4:2F47OFANFJGB421;4:GC;C2B4N;QMPZG<GN2H;13F=C;:GBCA2HbGBRS:2H69994N:23JN699x?sFTFb;CFGm������������c2<0BJ;=;<7q0����{;SS:;R̂?O;jJ;:>Q2g699@�D������5>xxxDe68L@@h���������Z=FX:2GC��������5\e���������h?e\1F==F2B_PE>695@D]:2n=;<];2b=;G=<2<;G:ANS2:cFBT;C}B KGA;122T ]FB4;:;<4c;BBG:q2:b2:G4F2B ]3=4;I:23b 2̂==X:24N;:< EEF<A=GFH;:

5CGR eCGR 5H2B4N hH2B4N 5R;G: eR;G: HGv��� KFBGBA; OGb< Q;d< PN2bbFBJ O2:; P;44FBJ< 2̂2=<���

Admin G/B/U Few R

3. Ranking Designs

• Like nominal, but ranked

• So use a bar chart

• And sort by value

• Put the item you want to call attention to at top or left

Admin G/B/U Few R

For Example

Courtesy of this site

Admin G/B/U Few R

4. Part-to-Whole Designs

• Comparison of shares

• Use simple bar

• Use stacked bar only when you want to compare acrosscategories

• So use a bar chart

• And sort by value

• Put the item you want to call attention to at top or left

Admin G/B/U Few R

For Example

Courtesy of this site

Admin G/B/U Few R

5. Deviation Designs

• Highlight differences across types

• Paired bars

• Doesn’t work too well with too many comparison categories

• Use stacked bar only when you want to compare acrosscategories

• More sophisticated (not in Few): scatterplot and compare to45 degree line

Admin G/B/U Few R

For Example

Admin G/B/U Few R

6. Distribution Designs

• Distributions can be continuous or by bin

• And you want to display one or many

• Use a bar chart

• Or a line chart

• Or a box plot – not too keen on these

Admin G/B/U Few R

For Example

Courtesy of this site

Admin G/B/U Few R

7. Correlation Designs

• Scatter chart

• Or scatter and trend line

• You can enhance scatter with color and weight variations

• Too many variations are not comprehensible

• No pic as you know this one

Admin G/B/U Few R

8. Geospatial Designs

• Map with

• Color fills

• Lines

• Much more on this later in the course

Admin G/B/U Few R

Bar Charts in R

Admin G/B/U Few R

Today’s Goals

• A few non-graph commands• ifelse• data.frame, c

• For graphing, via ggplot

• geom bar()• geom text(), geom label()• theme()

Admin G/B/U Few R

A Basic Programming Command: ifelse

data$var <- ifelse(test expression, [outcome if

true], [outcome if false])

• var

• Outcome is a variable in the dataframe data• Or something to do, instead of a variable

• test expression

• an expression that is evaluated, e.g. x > y?, a = b?

• After evaluation• if x > y , then you get the outcome if true – the second

element• if X < y , then you get the outcome if false – the third element• you can nest another ifelse in the third one

Admin G/B/U Few R

Get Started and Make Your Own Dataframe: data.frame

dfname <- data.frame(col1 =, col2 =, ...)

• data.frame creates a dataframe called dfname

• write column as name = c("e1", "e2", ... "en")

Make a dataframe

newframe <- data.frame(fruit = c("apples", "bananas","pomegranates"),

price.per.lb = c("2.49", "0.79","6"),

junk = rep(1,length(c("apples",

"bananas","pomegranates"))))

newframe

## fruit price.per.lb junk## 1 apples 2.49 1## 2 bananas 0.79 1## 3 pomegranates 6 1

Admin G/B/U Few R

Meryl’s Example from Last Class

A Dataframe from Last Class’s Bad Graph# load north korean datankd <- data.frame(year = c("2000","2001","2002","2003",

"2004","2005","2006","2007","2008","2009","2010","2011","2012","2013","2014","2015","2016","2017"),

defectors = c("0","0","1","0","0","0","0","0","2","0","1","0","3","0","0","1","1","4"))

nkd

## year defectors## 1 2000 0## 2 2001 0## 3 2002 1## 4 2003 0## 5 2004 0## 6 2005 0## 7 2006 0## 8 2007 0## 9 2008 2## 10 2009 0## 11 2010 1## 12 2011 0## 13 2012 3## 14 2013 0## 15 2014 0## 16 2015 1## 17 2016 1## 18 2017 4

Admin G/B/U Few R

And on to ggplot

For today we’re exploring

• geom bar

• geom text

• theme

Making a Bar Chart

# make a bar chartlibrary(ggplot2)ggplot(nkd, aes(x=defectors)) + geom_bar() +

labs(x = "annual number of defectors",y="number of years")

I call the ggplot libraryI a similar first part to last lecture:

I x axis is determined by quantity of defectorsI tell R we want a bar chart with geom_bar

I default is to count the total number of observations by type(defectors)

I labs() makes the graph comprehensible

What It Looks Like

0

3

6

9

0 1 2 3 4

annual number of defectors

num

ber

of y

ears

Other Options for geom_bar()

I if you want R to use the value in the dataframe, rather thancounting observations, use geom_bar(stat="identity")

I you can control aesthetics within the bar viageom_bar(aes(fill= [something])), useful for stackedgraphs

I you can weight the totalsI zillions more are available

geom_text() to Put Things on Your Chart

I puts variable value (maybe a fruit name) where you say basedon the value of another variable

I very powerful: need to set up data the right way to use thispower

geom_text() Example

Adding text to the chart with geom_text, telling R that themapping for labels is divisions$div.name.

ggplot(data = divisions, aes(x = division, y=ppl.by.cnty)) +geom_bar(stat = "identity") +ggtitle("counties by division using summarized data") +coord_flip() +labs(x="", y="people per county") +geom_text(mapping = aes(label=div.name))

What This Looks Like

New England

Mid−Atlantic

East North Central

West North Central

South Atlantic

East South Central

West South Central

Mountain

Pacific

2.5

5.0

7.5

0e+00 1e+05 2e+05 3e+05

people per county

counties by division using summarized data

Fixing the Previous

# make labels legibledivisions$nada <- c(rep(0, length(divisions$div.name)))ggplot(data = divisions, aes(x = division,

y=ppl.by.cnty)) +geom_bar(stat = "identity") +ggtitle("counties per capita by division

using summarized data") +coord_flip() +labs(x="", y="people per county") +geom_text(mapping = aes(y=nada, label=div.name),

hjust = 0)

I make a new variable that tells R where to put the name

How Does it Look?

New England

Mid−Atlantic

East North Central

West North Central

South Atlantic

East South Central

West South Central

Mountain

Pacific

2.5

5.0

7.5

0e+00 1e+05 2e+05 3e+05

people per county

counties per capita by division using summarized data

ggplot’s theme Commands

I a theme is a set of commands that standardize the look of thegraph

I ggplot has a built-in defaultI you can choose another defaultI or modify the themeI we’ll focus on the latter

Modifying the Default Theme

I there are > 60 different parts of the default theme, includingI axis.ticks.x()I legend.title()I legend.box.margin()

I see them all hereI in this class we mostly get rid of parts by adding the below to

the ggplot commandI theme(panel.grid.major = element_blank(),

panel.grid.minor = element_blank(),axis.ticks.y=element_blank())

Admin G/B/U Few R

Try Today’s Tutorial

• Pay attention to the output of each bit

• Go forth!

Admin G/B/U Few R

Next Lecture

• Turn in PS 4

• Read Few Chapter 6

• R Graphics Cookbook, Chapter 4

• Next policy brief deadline: April 2 for draft