See notes.R to follow along in RStudio.
Also see the R input from the lecture.
Check out these lecture notes!
Extra office hours today 5-7 pm in Shields 360.
Back to the NOAA earthquakes data!
q = read.delim("signif.txt")
names(q) = tolower(names(q))
quakes = qWhich country has the most earthquakes?
If we want to sort the table, use sort() after using table().
sort(table(q$country))
head(sort(table(q$country), decreasing = TRUE), 5)What if we want Chinese earthquakes from the last 200 years?
china200 = subset(q, country == "CHINA" & year >= 1816)How can we double-check?
# Check the year.
range(china200$year)
# 3 different ways to check the country.
sort(table(china200$country))
china200$country = factor(china200$country)
table(china200$country)
any(china200$country != "CHINA")Were any of these tsunamis?
colnames(china200)
levels(china200$flag_tsunami) = c("Quake", "Tsu")
head(china200$flag_tsunami, 20)What if we want a TRUE/FALSE for whether any are tsunamis?
is_tsu = china200$flag_tsunami == "Tsu"
any(is_tsu)
all(is_tsu)This data set is huge. How can we focus on just the part we're interested in?
Use the select parameter in the subset() function to narrow down the
columns.
# Cool columns we want to keep.
cool = c("flag_tsunami", "year", "focal_depth", "eq_primary", "country",
"latitude", "longitude")
cool_q = subset(q, select = cool)
head(cool_q)The second parameter (subset) in the subset() function controls rows,
while the third parameter (select) controls columns.
head(subset(china200, year >= 1916))
head(subset(china200, year >= 1916, select = cool))The order of function arguments doesn't matter if you write out the parameter names:
head(subset(china200, select = cool, subset = year >= 1916))What if we want to select rows and columns at the same time? What if we want the first 3 rows and first 3 columns?
Use [ to select [ROW, COLUMN]. Blank means select everything.
q[1, 1] # first row, first column
q[1, ] # first row, all columns
q[, 1] # all rows, first column
q[1:3, c(1, 5)]
q[1:3, ]
q[, 1:3]
head(q[1:3])
q[1] # all rows, first column (data frames are lists)
# What happens with negative indexes?
q[1:3, -1]Can we use column names instead of numbers?
q[c(1, 3), c("year", "country")]Can we use [ on a vector? Or is it just for
data frames?
a = c(5, 8, 2)
a[1]
a[2]
a[3]Does [ work like subset(), where we can use a
condition?
q[q$country == "CHINA", ]TRUE means "keep" and FALSE means "throw away".
q$country == "CHINA"It's easy to confuse the subset operator [
with the extraction operator [[. There are
two main differences:
-
[[extracts just one element, whereas[can get many. -
[[peels off lists, whereas[keeps them
For example:
q[[1, 1]]
a = list(6, 3, 1)
a[1]
a[[1]]
typeof(a[1])
typeof(a[[1]])In a picture:
a a[1] a[[1]]
+---+ +---+
| 6 | | 6 | 6
| 3 | +---+
| 1 |
+---+
Some functions are picky about vectors and lists.
The plot() function accepts a formula as the ## first argument. Formulas
in R are written ## y ~ x, and correspond to the plot axes.
q = quakes
plot(latitude ~ longitude, q)What's wrong with this plot?
It should have titles and axis labels. We could also add some color; we can
list R's built in colors with the colors() function.
colors()
plot(latitude ~ longitude, q, col = "plum", main = "Worldwide Earthquakes",
ylab = "Latitude", xlab = "Longitude")This plot doesn't necessarily need a legend, since there's only one kind of
point, but we could add one with the legend() function. The pch
parameter controls the point symbol used in the legend.
legend(-166.7, -36.5, legend = "Earthquake", pch = 1, col = "plum")Use the locator() function to figure out the coordinates for the legend's
top left corner.
locator(1)
plot(latitude ~ longitude, q, main = "Earthquakes", xlab = "Longitude",
ylab = "Latitude")How can we visualize which countries have the most earthquakes?
counts = sort(table(q$country))
barplot(counts)This is hard to read. We can swap the axes with the horiz parameter, and
rotate the labels with the las parameter. The latter is described in
# `?par`.
?par
barplot(counts, horiz = TRUE, las = 1)There's still too much on the plot. Let's cut down to the top 10.
barplot(tail(counts, 10), horiz = TRUE, las = 1, xlim = c(0, 1000),
main = "Earthquakes", ylab = "Country", xlab = "Number of Earthquakes")Does the width of the bars tell us anything?
Not really, so let's dump them.
dotchart(tail(counts, 10), xlim = c(0, 1000))How can we show just the top 20%?
top20 = length(unique(q$country)) * .2
top20 = tail(counts, top20)
dotchart(top20)The dotchart() function is not happy that top20 has class array
instead of being a vector. We can fix it with the vector creation function
c().
class(top20)
top20 = c(top20)What's still missing?
Every plot should have a title and labels.
dotchart(top10, xlim = c(0, 1000), main = "Top Earthquakes",
xlab = "Number of Quakes")How can we separate regular quakes and tsunami?
tsu = table(q$country, q$flag_tsunami)Make the table into a data frame so we can use $ to extract columns.
tsu = data.frame(Quake = tsu[, 1], Tsu = tsu[, 2])
head(tsu)How can we sort 2-dimensional data? The sort() function won't work.
Use the subset operator [ to reorganize rows.
tsu[1:3, ]
tsu[c(2, 3, 1), ]We also need a function that can tell us which row goes first, which goes
second, and so on. That's the order() function.
ord = order(tsu$Quake + tsu$Tsu)
tsu[ord, ]