STAT545A hw04: Visualize a quantitative variable with lattice

Rebecca Johnston

Here I will be exploring some features of lattice, a multivariate data visualization package in R. I will use the gapminder data to provide examples of the graphical output from this package.

Load required packages and gapminder data
How is life expectancy changing over time on different continents?
Depict the maximum and minimum of GDP per capita for all continents

Load required packages and gapminder data

To begin, load the required packages for this session:

library(lattice)
library(plyr)
library(xtable)

Load the gapminder data and check it loaded correctly:

gdURL <- "http://www.stat.ubc.ca/~jenny/notOcto/STAT545A/examples/gapminder/data/gapminderDataFiveYear.txt"
gDat <- read.delim(file = gdURL)
str(gDat)

## 'data.frame':    1704 obs. of  6 variables:
##  $ country  : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
##  $ pop      : num  8425333 9240934 10267083 11537966 13079460 ...
##  $ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
##  $ lifeExp  : num  28.8 30.3 32 34 36.1 ...
##  $ gdpPercap: num  779 821 853 836 740 ...

table(gDat$continent, gDat$year)

##           
##            1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 2002 2007
##   Africa     52   52   52   52   52   52   52   52   52   52   52   52
##   Americas   25   25   25   25   25   25   25   25   25   25   25   25
##   Asia       33   33   33   33   33   33   33   33   33   33   33   33
##   Europe     30   30   30   30   30   30   30   30   30   30   30   30
##   Oceania     2    2    2    2    2    2    2    2    2    2    2    2

Jenny's tip: Drop Oceania from further analyses straight away, as it only contains data for 2 countries.
To do this, create a subset of gDat without the continent Oceania, by using the function droplevels:

iDat <- droplevels(subset(gDat, continent != "Oceania"))
table(iDat$continent)

## 
##   Africa Americas     Asia   Europe 
##      624      300      396      360

How is life expectancy changing over time on different continents?

I have chosen to create graphs to accompany code created by Yumian Hu. Not only did Yumian answer every single question from hw03 (so impressive!), she answered each question with clean code. I even learnt a few tricks from her .rmd file to create links within this very page. Thanks Yumian!

Get a data.frame with 1 + 5 = 6 variables: year and then one variable per continent, giving mean or median life expectancy. Use daply to obtain data in “wide” format:

meanLife <- as.data.frame(daply(iDat, ~year + continent, summarize, mean(lifeExp)))  # Code adapted from Yumian Hu
print(xtable(meanLife), type = "html", include.rownames = TRUE)

	Africa	Americas	Asia	Europe
1952	39.14	53.28	46.31	64.41
1957	41.27	55.96	49.32	66.70
1962	43.32	58.40	51.56	68.54
1967	45.33	60.41	54.66	69.74
1972	47.45	62.39	57.32	70.78
1977	49.58	64.39	59.61	71.94
1982	51.59	66.23	62.62	72.81
1987	53.34	68.09	64.85	73.64
1992	53.63	69.57	66.54	74.44
1997	53.60	71.15	68.02	75.51
2002	53.33	72.42	69.23	76.70
2007	54.81	73.61	70.73	77.65

It is FAR more elegant to display this data in “wide” format vs. “tall” format. However! I quickly discovered that using “wide” format as input for lattice was particularly difficult!
Therefore, to plot data using lattice I will use the “tall” format to easily call upon groups and variables. Now I understand the importance of being able to switch between tall and wide data!!!
Use ddply to obtain data in “tall” format:

meanLife <- ddply(iDat, ~continent + year, summarize, avgLifeExp = mean(lifeExp))
head(meanLife)  # Too tall to display everything!

##   continent year avgLifeExp
## 1    Africa 1952      39.14
## 2    Africa 1957      41.27
## 3    Africa 1962      43.32
## 4    Africa 1967      45.33
## 5    Africa 1972      47.45
## 6    Africa 1977      49.58

Create a scatter plot of the results using xyplot:
- 1st argument = y vs. x
- 2nd argument = the data (doesn't go first like in plyr)
- 3rd argument = how the data should be grouped
- 4th argument = a key should be produced

xyplot(avgLifeExp ~ year, meanLife, groups = continent, auto.key = TRUE)

plot of chunk unnamed-chunk-6

Notice that the order of the continents within the figure legend does not correspond to the order of the data within the graph.
- I would prefer that the key lists the continents Europe, Americas, Asia, then Africa. I would also prefer the placement of the key to the right for ease of viewing.
These (aesthetic) issues can be rectified by changing a few details within the auto.key argument. I have also added lines of regression by calling for type = “r” to help visualize trends in average life expectancy over time by continent:

xyplot(avgLifeExp ~ year, meanLife, groups = continent, grid = "h", type = c("p", 
    "r"), auto.key = list(space = "right", reverse.rows = TRUE))

plot of chunk unnamed-chunk-7

We can easily see that the mean life expectancy for all continents is increasing with time. However the rate (slopes) at which life expectancy values are increasing are not equal across the four continents.

Depict the maximum and minimum of GDP per capita for all continents

I will begin with code adapted from Yumian Hu, but I will then need to work the year variable into the visual display.

GDPpercap = ddply(iDat, year ~ continent, summarize, minGDP = min(gdpPercap), 
    maxGDP = max(gdpPercap))  # Code adapted from Yumian Hu
head(GDPpercap)  # Too long to show it all

##   year continent minGDP maxGDP
## 1 1952    Africa  298.8   4725
## 2 1952  Americas 1397.7  13990
## 3 1952      Asia  331.0 108382
## 4 1952    Europe  973.5  14734
## 5 1957    Africa  336.0   5487
## 6 1957  Americas 1544.4  14847

print(xtable(GDPpercap), type = "html", include.rownames = FALSE)

year	continent	minGDP	maxGDP
1952	Africa	298.85	4725.30
1952	Americas	1397.72	13990.48
1952	Asia	331.00	108382.35
1952	Europe	973.53	14734.23
1957	Africa	336.00	5487.10
1957	Americas	1544.40	14847.13
1957	Asia	350.00	113523.13
1957	Europe	1353.99	17909.49
1962	Africa	355.20	6757.03
1962	Americas	1662.14	16173.15
1962	Asia	388.00	95458.11
1962	Europe	1709.68	20431.09
1967	Africa	412.98	18772.75
1967	Americas	1452.06	19530.37
1967	Asia	349.00	80894.88
1967	Europe	2172.35	22966.14
1972	Africa	464.10	21011.50
1972	Americas	1654.46	21806.04
1972	Asia	357.00	109347.87
1972	Europe	2860.17	27195.11
1977	Africa	502.32	21951.21
1977	Americas	1874.30	24072.63
1977	Asia	371.00	59265.48
1977	Europe	3528.48	26982.29
1982	Africa	462.21	17364.28
1982	Americas	2011.16	25009.56
1982	Asia	424.00	33693.18
1982	Europe	3630.88	28397.72
1987	Africa	389.88	11864.41
1987	Americas	1823.02	29884.35
1987	Asia	385.00	28118.43
1987	Europe	3738.93	31540.97
1992	Africa	410.90	13522.16
1992	Americas	1456.31	32003.93
1992	Asia	347.00	34932.92
1992	Europe	2497.44	33965.66
1997	Africa	312.19	14722.84
1997	Americas	1341.73	35767.43
1997	Asia	415.00	40300.62
1997	Europe	3193.05	41283.16
2002	Africa	241.17	12521.71
2002	Americas	1270.36	39097.10
2002	Asia	611.00	36023.11
2002	Europe	4604.21	44683.98
2007	Africa	277.55	13206.48
2007	Americas	1201.64	42951.65
2007	Asia	944.00	47306.99
2007	Europe	5937.03	49357.19

I can simply generate min and max GDPpercap per year on separate graphs:

xyplot(minGDP ~ year, GDPpercap, groups = continent, auto.key = TRUE)

plot of chunk unnamed-chunk-10

xyplot(maxGDP ~ year, GDPpercap, groups = continent, auto.key = TRUE)

plot of chunk unnamed-chunk-10

The trick is to generate them on the same graph!
One way to go about it is to write a function that creates a new column in the dataframe GDPpercap to specify whether the value for each continent is min or max, then call upon ddply to run the function:

minmax <- function(x) {
    stat = c("Min", "Max")
    gdpPerCap = c(min(x$gdpPercap), max(x$gdpPercap))
    data.frame(stat, gdpPerCap)
}
gdpYear = ddply(iDat, year ~ continent, minmax)
head(gdpYear)

##   year continent stat gdpPerCap
## 1 1952    Africa  Min     298.8
## 2 1952    Africa  Max    4725.3
## 3 1952  Americas  Min    1397.7
## 4 1952  Americas  Max   13990.5
## 5 1952      Asia  Min     331.0
## 6 1952      Asia  Max  108382.4

str(gdpYear)

## 'data.frame':    96 obs. of  4 variables:
##  $ year     : int  1952 1952 1952 1952 1952 1952 1952 1952 1957 1957 ...
##  $ continent: Factor w/ 4 levels "Africa","Americas",..: 1 1 2 2 3 3 4 4 1 1 ...
##  $ stat     : Factor w/ 2 levels "Max","Min": 2 1 2 1 2 1 2 1 2 1 ...
##  $ gdpPerCap: num  299 4725 1398 13990 331 ...

Now to plot the results:

xyplot(gdpPerCap ~ year | stat, gdpYear, groups = continent, auto.key = TRUE)

plot of chunk unnamed-chunk-12

The “min” gdpPerCap panel does not provide much information. I think it would be best to set an independent y scale for this plot to visualize whether there are differences in values between the continents.
I also find it distracting that data points are overlapping in the Max gdpPerCap panel. We should join the points.
Let's replot the graph with a few extra arguments:
- scales = list(y = list(relation = “free”)) = separate y scales
- type = c(“p”, “l”) = plot points (p) connected by lines (l)

xyplot(gdpPerCap ~ year | stat, gdpYear, groups = continent, type = c("p", "l"), 
    scales = list(y = list(relation = "free")), auto.key = list(columns = nlevels(gdpYear$continent)))

plot of chunk unnamed-chunk-13

I can swap the order of the panels using the function reorder:

gdpYear <- within(gdpYear, stat <- reorder(stat, gdpPerCap, FUN = min))
xyplot(gdpPerCap ~ year | stat, gdpYear, groups = continent, type = c("p", "l"), 
    scales = list(y = list(relation = "free")), auto.key = list(columns = nlevels(gdpYear$continent)))

plot of chunk unnamed-chunk-14

We can observe that the minimum GdpPerCap value for Europe has been increasing at a relatively fast rate, Whereas the Americas, Asia and Africa appear to have a steady minimum GdpPerCap value. Although Asia looks to be picking up over the last decade or so.
On the other hand, it is striking that the maximum GdpPerCap value dropped dramatically for Asia from about 1950 - 1985, but it has been increasing ever since, albeit at a slower rate. Americas and Europe have a very similar trend, with a steady growth in the maxGdpPerCap with time.