1
00:00:09,490 --> 00:00:13,140
In this video, we're
going to look at scales.

2
00:00:13,140 --> 00:00:15,260
This first plot shows
the average height

3
00:00:15,260 --> 00:00:18,980
of a 21-year-old
male in centimeters.

4
00:00:18,980 --> 00:00:25,300
The x-axis is time, starting
in 1871, and ending in 1975.

5
00:00:25,300 --> 00:00:27,270
Each person
represents the height,

6
00:00:27,270 --> 00:00:30,320
at a different point in time,
and the points are evenly

7
00:00:30,320 --> 00:00:34,530
spaced in time, so
the x-axis is OK.

8
00:00:34,530 --> 00:00:39,820
The y-axis ranges from just
under 160 to 180 centimeters,

9
00:00:39,820 --> 00:00:44,030
which isn't inherently bad,
but does overstate the change.

10
00:00:44,030 --> 00:00:46,210
The real problem is the bars.

11
00:00:46,210 --> 00:00:47,590
If it was accurate,
we would only

12
00:00:47,590 --> 00:00:49,940
really see the heads of
the men, but instead we

13
00:00:49,940 --> 00:00:52,770
see their whole bodies, making
it seem as if people have not

14
00:00:52,770 --> 00:00:55,320
only doubled in height, but
they've also double in width.

15
00:00:59,440 --> 00:01:02,590
This next plot also
has issues with scale.

16
00:01:02,590 --> 00:01:05,850
The total range of
the plot is 8% to 10%,

17
00:01:05,850 --> 00:01:11,360
although all the numbers fall
in the range of 8.6% to 9.2%.

18
00:01:11,360 --> 00:01:14,490
If we plotted the y-axis
on a 0% to 10% scale,

19
00:01:14,490 --> 00:01:16,520
the conclusion would be
that nothing is really

20
00:01:16,520 --> 00:01:18,520
changing at all.

21
00:01:18,520 --> 00:01:21,190
The last point in the chart
is at the wrong height,

22
00:01:21,190 --> 00:01:23,810
and the size of the markers
makes the relative locations

23
00:01:23,810 --> 00:01:25,860
hard to distinguish.

24
00:01:25,860 --> 00:01:31,770
Also notice that the gap between
9.0% and the 8.9% markers

25
00:01:31,770 --> 00:01:36,820
on the far left side, and
the 8.9% and 8.8% markers,

26
00:01:36,820 --> 00:01:38,789
have a different gap.

27
00:01:38,789 --> 00:01:41,550
This plot shows the relative
breakdown of teachers

28
00:01:41,550 --> 00:01:44,430
by race in a certain
teaching program.

29
00:01:44,430 --> 00:01:47,560
The Caucasian bar is truncated,
which is a risky choice,

30
00:01:47,560 --> 00:01:50,759
but could be appropriate
in some situations.

31
00:01:50,759 --> 00:01:53,620
A much bigger problem is that
the scale of each blue bar

32
00:01:53,620 --> 00:01:55,500
is entirely different.

33
00:01:55,500 --> 00:01:58,450
For example, the Native
American bar is about a third

34
00:01:58,450 --> 00:02:00,770
of the length of the
African American bar,

35
00:02:00,770 --> 00:02:03,680
but there are more than 10
times as many African Americans

36
00:02:03,680 --> 00:02:06,170
in this program as
Native Americans.

37
00:02:06,170 --> 00:02:09,860
In fact, visually, this plot
is completely meaningless.

38
00:02:09,860 --> 00:02:11,960
The only useful thing
about it is the numbers.

39
00:02:11,960 --> 00:02:14,750
But even there, there
is a bit of confusion,

40
00:02:14,750 --> 00:02:18,020
as Native Americans are
given to one decimal place,

41
00:02:18,020 --> 00:02:19,470
but the others are rounded.

42
00:02:19,470 --> 00:02:21,820
Which when combined with
the confusing scales,

43
00:02:21,820 --> 00:02:23,800
casts doubt on the
correctness of the numbers.

44
00:02:27,950 --> 00:02:31,300
Here is a before and
after of the same data.

45
00:02:31,300 --> 00:02:33,930
On the left, we see
the US military expense

46
00:02:33,930 --> 00:02:37,530
in the right axis, and troop
count on the left axis.

47
00:02:37,530 --> 00:02:40,890
Both the line and bar
plots are individually OK,

48
00:02:40,890 --> 00:02:43,280
but the combination
is misleading.

49
00:02:43,280 --> 00:02:46,390
Because you have mixed two
units, dollars and people,

50
00:02:46,390 --> 00:02:48,060
there is a false
impression of some sort

51
00:02:48,060 --> 00:02:52,850
of crossover point in
1995 that does not exist.

52
00:02:52,850 --> 00:02:56,650
On the right is the same data
presented in a different way.

53
00:02:56,650 --> 00:02:58,430
We now have troops
on the x-axis,

54
00:02:58,430 --> 00:03:00,710
and dollars on the y-axis.

55
00:03:00,710 --> 00:03:02,740
The line moves through
time now, allowing

56
00:03:02,740 --> 00:03:05,990
us to see when moments of change
occurred, such as decreases

57
00:03:05,990 --> 00:03:09,510
in troop count, through the
90s, at the end of the Cold War,

58
00:03:09,510 --> 00:03:11,690
the increase in
spending of the 2000s,

59
00:03:11,690 --> 00:03:14,850
and the recent decreases
in military spending.

60
00:03:14,850 --> 00:03:17,500
The final visualization
I want to show you today

61
00:03:17,500 --> 00:03:20,280
is all about the different
types of household.

62
00:03:20,280 --> 00:03:22,280
The US Census
Bureau periodically

63
00:03:22,280 --> 00:03:25,900
determines how many households
are comprised, for example,

64
00:03:25,900 --> 00:03:28,560
of married couples with
and without children,

65
00:03:28,560 --> 00:03:32,320
people living alone, and so on.

66
00:03:32,320 --> 00:03:35,730
First of all, I'm not saying
this is a bad visualization.

67
00:03:35,730 --> 00:03:38,280
In fact, if we are interested
in the relative share

68
00:03:38,280 --> 00:03:41,370
of each type of household
in a particular year,

69
00:03:41,370 --> 00:03:43,620
it's actually pretty good.

70
00:03:43,620 --> 00:03:45,630
However, if what
we're interested in

71
00:03:45,630 --> 00:03:47,940
is the rates of change
across the years,

72
00:03:47,940 --> 00:03:50,010
this is next to useless.

73
00:03:50,010 --> 00:03:53,790
The key problem is that the
x-axis is completely off.

74
00:03:53,790 --> 00:03:56,690
The gap between the first
two columns is 10 years,

75
00:03:56,690 --> 00:03:58,590
but the gap between
the last two columns

76
00:03:58,590 --> 00:04:01,090
is only 2 years, meaning
that the rates are

77
00:04:01,090 --> 00:04:03,150
hard to read from this.

78
00:04:03,150 --> 00:04:05,510
If we're not interested
in the rates of changes,

79
00:04:05,510 --> 00:04:09,920
but just want to compare two
years at a time, it's not bad,

80
00:04:09,920 --> 00:04:12,070
but it's not easy either.

81
00:04:12,070 --> 00:04:15,170
Try comparing 1970
married without children

82
00:04:15,170 --> 00:04:18,100
to 2010 married
without children,

83
00:04:18,100 --> 00:04:20,250
without looking at the numbers.

84
00:04:20,250 --> 00:04:23,100
Can you tell if it
has grown or shrunk?

85
00:04:23,100 --> 00:04:25,490
Finally, and more
generally, this chart

86
00:04:25,490 --> 00:04:27,500
shows relative numbers.

87
00:04:27,500 --> 00:04:29,380
If you look at
absolute numbers, we

88
00:04:29,380 --> 00:04:32,530
might find the total number of
couples married with children

89
00:04:32,530 --> 00:04:35,560
is actually constant, but the
number of other households

90
00:04:35,560 --> 00:04:38,350
has increased.

91
00:04:38,350 --> 00:04:40,140
We are now going
to change into R

92
00:04:40,140 --> 00:04:43,350
to try plotting this
data as a line chart.