1
00:00:04,500 --> 00:00:08,420
First, let's make sure
we have ggplot2 loaded.

2
00:00:08,420 --> 00:00:10,110
So library(ggplot2).

3
00:00:16,420 --> 00:00:20,460
Now let's load our data frame,
which is in households.csv.

4
00:00:20,460 --> 00:00:22,880
So read.csv(households.csv).

5
00:00:31,190 --> 00:00:36,180
If we look at the
structure of households,

6
00:00:36,180 --> 00:00:39,020
we see that there is
a year column and then

7
00:00:39,020 --> 00:00:44,220
six other columns for each of
the different household types.

8
00:00:44,220 --> 00:00:46,740
So this is actually
a problem for us.

9
00:00:46,740 --> 00:00:49,190
Given this structure
of a data frame,

10
00:00:49,190 --> 00:00:50,840
what would we put
in the aesthetic

11
00:00:50,840 --> 00:00:54,370
for our ggplot command?

12
00:00:54,370 --> 00:00:56,710
It's not obvious,
to me at least,

13
00:00:56,710 --> 00:00:59,970
and in fact, I don't think
it's really possible.

14
00:00:59,970 --> 00:01:02,040
The reason is that
ggplot needs it

15
00:01:02,040 --> 00:01:08,750
in the form of: year,
group, and fraction.

16
00:01:11,330 --> 00:01:14,150
The solution is to
use the melt function

17
00:01:14,150 --> 00:01:16,720
from the reshape package.

18
00:01:16,720 --> 00:01:19,920
Melt will take a 2-dimensional
data frame like ours,

19
00:01:19,920 --> 00:01:24,550
and convert it into exactly the
right form we need for ggplot2.

20
00:01:24,550 --> 00:01:28,490
So first, let's load reshape2
-- library(reshape2).

21
00:01:31,820 --> 00:01:35,229
Now, let's look at the first two
columns of our households data

22
00:01:35,229 --> 00:01:37,880
frame -- households[,1:2].

23
00:01:45,750 --> 00:01:49,610
So there's a Year and a
MarriedWChild for each year.

24
00:01:49,610 --> 00:01:53,390
Now, let's look at the first few
rows of our melted households

25
00:01:53,390 --> 00:01:54,320
data frame.

26
00:01:54,320 --> 00:01:55,860
So head(melt(households,
id="Year")).

27
00:02:09,990 --> 00:02:10,930
And there you have it.

28
00:02:10,930 --> 00:02:13,000
So, basically,
what's happened is

29
00:02:13,000 --> 00:02:16,730
that each value of
MarriedWChild has

30
00:02:16,730 --> 00:02:19,870
turned into its own row
in the new data frame.

31
00:02:22,810 --> 00:02:24,610
To make it more clear,
perhaps, let's look

32
00:02:24,610 --> 00:02:26,610
at the first three
columns of households.

33
00:02:32,890 --> 00:02:35,250
Now we have MarriedWOChild.

34
00:02:35,250 --> 00:02:39,500
Now let's look at, instead
of just the first rows

35
00:02:39,500 --> 00:02:43,290
of our melted data frame, let's
look at the first 10 rows.

36
00:02:43,290 --> 00:02:46,070
So rows 1 to 10, all columns.

37
00:02:53,420 --> 00:02:55,750
There we go.

38
00:02:55,750 --> 00:03:01,110
So there you can see the
eight values of MarriedWChild,

39
00:03:01,110 --> 00:03:03,810
and the first two values
of MarriedWOChild.

40
00:03:03,810 --> 00:03:09,020
So there's that 30.3 up
there, gone down to 30.3 here,

41
00:03:09,020 --> 00:03:11,540
29.9 gone to down here.

42
00:03:11,540 --> 00:03:13,280
So every value in
our data frame now

43
00:03:13,280 --> 00:03:17,470
corresponds to a new row
in our melted data frame,

44
00:03:17,470 --> 00:03:20,070
which is exactly what
we need for ggplot.

45
00:03:20,070 --> 00:03:26,060
So let's try plotting this
melted data frame -- ggplot,

46
00:03:26,060 --> 00:03:33,430
melt, households, using the
Year column as an id column,

47
00:03:33,430 --> 00:03:35,670
the key column.

48
00:03:35,670 --> 00:03:41,910
Our aesthetic is going to be
now to use Year on the x-axis.

49
00:03:41,910 --> 00:03:45,260
Our y-axis will be the value
column of our melted data

50
00:03:45,260 --> 00:03:47,970
frame.

51
00:03:47,970 --> 00:03:50,579
And the color of
the line will depend

52
00:03:50,579 --> 00:03:53,850
on the group, which is called
variable in the melted data

53
00:03:53,850 --> 00:03:54,350
frame.

54
00:04:00,080 --> 00:04:01,430
So that's our aesthetic.

55
00:04:01,430 --> 00:04:05,220
Our geometry is
going to be lines.

56
00:04:05,220 --> 00:04:07,610
And I want to make the
lines a little bit thicker.

57
00:04:07,610 --> 00:04:12,380
So let's say line size is 2.

58
00:04:12,380 --> 00:04:16,060
And I also want to have
points for each year

59
00:04:16,060 --> 00:04:18,470
in the data frame, so I'm
going to have lines and points.

60
00:04:18,470 --> 00:04:20,820
So geom_point.

61
00:04:20,820 --> 00:04:23,270
And I'm going to make the size
of these a little bit bigger

62
00:04:23,270 --> 00:04:26,010
than normal too, size = 5.

63
00:04:26,010 --> 00:04:28,140
And we should put
a y-axis label.

64
00:04:28,140 --> 00:04:29,640
So ylab("Percentage
of Households").

65
00:04:37,710 --> 00:04:39,970
And there you go.

66
00:04:39,970 --> 00:04:42,350
Now, this is actually
quite interesting

67
00:04:42,350 --> 00:04:45,670
when we compare it back to the
chart we had in the slides.

68
00:04:45,670 --> 00:04:48,159
Now you can see just how
quickly MarriedWChild

69
00:04:48,159 --> 00:04:49,760
is decreasing as
a relative share.

70
00:04:49,760 --> 00:04:53,960
You can also more clearly
see that MarriedWOChild

71
00:04:53,960 --> 00:04:57,490
is pretty much flat, and that
the differences being made up

72
00:04:57,490 --> 00:04:59,790
by the other four
types of households

73
00:04:59,790 --> 00:05:01,430
is steadily increasing
over the years.

74
00:05:03,990 --> 00:05:05,820
So there you have
it, the same data,

75
00:05:05,820 --> 00:05:07,780
plotted in two different ways.

76
00:05:07,780 --> 00:05:11,410
Now, I'm not saying one of these
is better than the other one.

77
00:05:11,410 --> 00:05:13,940
For example, if I want to
compare inside a given year,

78
00:05:13,940 --> 00:05:18,330
say 1970, it's not the most
easy thing, at a glance,

79
00:05:18,330 --> 00:05:22,370
to see just how much of
a total hundred percent

80
00:05:22,370 --> 00:05:24,280
is taken up by each.

81
00:05:24,280 --> 00:05:26,980
But if I want to see across
years, it's far superior.

82
00:05:26,980 --> 00:05:29,980
And I can clearly see that the
last data point is pretty much

83
00:05:29,980 --> 00:05:31,730
right next to the
second to last data

84
00:05:31,730 --> 00:05:33,150
point, which is
something that was

85
00:05:33,150 --> 00:05:36,690
hard to tell with the
other visualization.

86
00:05:36,690 --> 00:05:38,970
So I hope this has made
you think a little bit more

87
00:05:38,970 --> 00:05:41,560
about the different ways
you can plot the same data.

88
00:05:41,560 --> 00:05:43,690
And hopefully improved
your ggplot2 skills

89
00:05:43,690 --> 00:05:45,270
a little bit more.

90
00:05:45,270 --> 00:05:47,140
Thanks for watching.