1
00:00:01,420 --> 00:00:04,390
A random variable can take
different numerical values

2
00:00:04,390 --> 00:00:06,890
depending on the outcome
of the experiment.

3
00:00:06,890 --> 00:00:09,800
Some outcomes are more likely
than others, and similarly

4
00:00:09,800 --> 00:00:13,130
some of the possible numerical
values of a random variable

5
00:00:13,130 --> 00:00:15,480
will be more likely
than others.

6
00:00:15,480 --> 00:00:18,500
We restrict ourselves to
discrete random variables, and

7
00:00:18,500 --> 00:00:22,060
we will describe these relative
likelihoods in terms

8
00:00:22,060 --> 00:00:26,080
of the so-called probability
mass function, or PMF for

9
00:00:26,080 --> 00:00:29,440
short, which gives the
probability of the different

10
00:00:29,440 --> 00:00:31,530
possible numerical values.

11
00:00:31,530 --> 00:00:35,480
The PMF is also sometimes called
the probability law or

12
00:00:35,480 --> 00:00:39,590
the probability distribution of
a discrete random variable.

13
00:00:39,590 --> 00:00:43,090
Let me illustrate the idea in
terms of a simple example.

14
00:00:43,090 --> 00:00:45,160
We have a probabilistic
experiment with

15
00:00:45,160 --> 00:00:47,380
four possible outcomes.

16
00:00:47,380 --> 00:00:52,600
We also have a probability
law on the sample space.

17
00:00:52,600 --> 00:00:57,170
And to keep things simple, we
assume that all four outcomes

18
00:00:57,170 --> 00:01:00,610
in our sample space are
equally likely.

19
00:01:00,610 --> 00:01:03,450
We then introduce a random
variable that associates a

20
00:01:03,450 --> 00:01:05,800
number with each possible
outcome as

21
00:01:05,800 --> 00:01:07,630
shown in this diagram.

22
00:01:07,630 --> 00:01:12,310
The random variable,
X, can take one of

23
00:01:12,310 --> 00:01:14,030
three possible values--

24
00:01:14,030 --> 00:01:16,510
namely 3, 4, or 5.

25
00:01:16,510 --> 00:01:19,810
Let us focus on one
of those numbers--

26
00:01:19,810 --> 00:01:22,610
let's say the number 5.

27
00:01:22,610 --> 00:01:28,100
So let us focus on x
being equal to 5.

28
00:01:28,100 --> 00:01:34,990
We can think of the event
that X is equal to 5.

29
00:01:34,990 --> 00:01:36,890
Which event is this?

30
00:01:36,890 --> 00:01:40,090
This is the event that the
outcome of the experiment led

31
00:01:40,090 --> 00:01:44,310
to the random variable
taking a value of 5.

32
00:01:44,310 --> 00:01:47,470
So it is this particular event
which consists of two

33
00:01:47,470 --> 00:01:50,729
elements, namely a and b.

34
00:01:50,729 --> 00:01:55,580
More formally, the event that
we're talking about is the set

35
00:01:55,580 --> 00:02:01,790
of all outcomes for which the
value, the numerical value of

36
00:02:01,790 --> 00:02:05,200
our random variable, which is
a function of the outcome,

37
00:02:05,200 --> 00:02:08,770
that numerical value happens
to be equal to 5.

38
00:02:08,770 --> 00:02:11,880
And in this example it is a set

39
00:02:11,880 --> 00:02:13,410
consisting of two elements.

40
00:02:13,410 --> 00:02:15,460
It's a subset of the
sample space.

41
00:02:15,460 --> 00:02:17,750
So it is an event.

42
00:02:17,750 --> 00:02:19,700
And it has a probability.

43
00:02:19,700 --> 00:02:21,510
And that probability we will be

44
00:02:21,510 --> 00:02:25,500
denoting with this notation.

45
00:02:25,500 --> 00:02:30,160
And in our case this probability
is equal to 1/2.

46
00:02:30,160 --> 00:02:34,310
Because we have two outcomes,
each one has probability 1/4.

47
00:02:34,310 --> 00:02:38,890
The probability of this
event is equal to 1/2.

48
00:02:38,890 --> 00:02:45,290
More generally, we will be using
this notation to denote

49
00:02:45,290 --> 00:02:50,730
the probability of the event
that the random variable, X ,

50
00:02:50,730 --> 00:02:53,900
takes on a particular
value, x.

51
00:02:53,900 --> 00:02:57,130
This is just a piece of
notation, not a new concept.

52
00:02:57,130 --> 00:02:59,740
We're dealing with a
probability, and we indicate

53
00:02:59,740 --> 00:03:04,350
it using this particular
notation.

54
00:03:04,350 --> 00:03:08,550
More formally, the probability
that we're dealing with is the

55
00:03:08,550 --> 00:03:12,950
probability, the total
probability, of all outcomes

56
00:03:12,950 --> 00:03:17,360
for which the numerical value of
our random variable is this

57
00:03:17,360 --> 00:03:20,320
particular number, x.

58
00:03:20,320 --> 00:03:21,960
A few things to notice.

59
00:03:21,960 --> 00:03:28,780
We use a subscript, X, to
indicate which random variable

60
00:03:28,780 --> 00:03:30,230
we're talking about.

61
00:03:30,230 --> 00:03:32,510
This will be useful
if we have several

62
00:03:32,510 --> 00:03:34,300
random variables involved.

63
00:03:34,300 --> 00:03:37,060
For example, if we have another
random variable on the

64
00:03:37,060 --> 00:03:42,890
same sample space, Y, then it
would have its own probability

65
00:03:42,890 --> 00:03:48,079
mass function which would be
denoted with this particular

66
00:03:48,079 --> 00:03:49,329
notation here.

67
00:03:52,160 --> 00:03:59,470
The argument of the PMF, which
is x, ranges over the possible

68
00:03:59,470 --> 00:04:04,750
values of the random variable,
X. So in this sense, here

69
00:04:04,750 --> 00:04:08,020
we're really dealing
with a function.

70
00:04:08,020 --> 00:04:12,200
A function that we could
denote just by p with a

71
00:04:12,200 --> 00:04:13,310
subscript x.

72
00:04:13,310 --> 00:04:16,170
This is a function as opposed
to the specific

73
00:04:16,170 --> 00:04:17,839
values of this function.

74
00:04:17,839 --> 00:04:21,350
And we can produce plots
of this function.

75
00:04:21,350 --> 00:04:25,050
In this particular example that
we're dealing with, the

76
00:04:25,050 --> 00:04:29,890
interesting values of
x are 3, 4, and 5.

77
00:04:29,890 --> 00:04:33,640
And the associated probabilities
are the value of

78
00:04:33,640 --> 00:04:39,790
5 is obtained with probability
1/2, the value of 4--

79
00:04:39,790 --> 00:04:45,930
this is the event that the
outcome is c, which has

80
00:04:45,930 --> 00:04:48,070
probability 1/4.

81
00:04:48,070 --> 00:04:52,960
And the value of 3 is also
obtained with probability 1/4

82
00:04:52,960 --> 00:04:56,420
because the value of 3 is
obtained when the outcome is

83
00:04:56,420 --> 00:04:59,740
d, and that outcome has
probability 1/4.

84
00:04:59,740 --> 00:05:03,390
So the probability mass function
is a function of an

85
00:05:03,390 --> 00:05:05,480
argument x.

86
00:05:05,480 --> 00:05:08,920
And for any x, it specifies
the probability that the

87
00:05:08,920 --> 00:05:12,930
random variable takes on
this particular value.

88
00:05:12,930 --> 00:05:15,710
A few more things to notice.

89
00:05:15,710 --> 00:05:19,150
The probability mass function is
always non-negative, since

90
00:05:19,150 --> 00:05:20,750
we're talking about
probabilities and

91
00:05:20,750 --> 00:05:23,930
probabilities are always
non-negative.

92
00:05:23,930 --> 00:05:28,350
In addition, since the total
probability of all outcomes is

93
00:05:28,350 --> 00:05:32,590
equal to 1, the probabilities
of the different possible

94
00:05:32,590 --> 00:05:37,330
values of the random variable
should also add to 1.

95
00:05:37,330 --> 00:05:42,040
So when you add over all
possible values of x, the sum

96
00:05:42,040 --> 00:05:43,790
of the associated probabilities

97
00:05:43,790 --> 00:05:45,740
should be equal to 1.

98
00:05:45,740 --> 00:05:52,690
In terms of our picture, the
event that x is equal to 3,

99
00:05:52,690 --> 00:05:57,200
which is this subset of the
sample space, the event that x

100
00:05:57,200 --> 00:06:03,520
is equal to 4, which is this
subset of the sample space,

101
00:06:03,520 --> 00:06:06,990
and the event that x is equal to
5, which is this subset of

102
00:06:06,990 --> 00:06:08,470
the sample space.

103
00:06:08,470 --> 00:06:09,840
These three events--

104
00:06:09,840 --> 00:06:12,080
the red, green, and blue--

105
00:06:12,080 --> 00:06:14,940
they are disjoint, and together
they cover the entire

106
00:06:14,940 --> 00:06:15,820
sample space.

107
00:06:15,820 --> 00:06:18,940
So their probabilities
should add to 1.

108
00:06:18,940 --> 00:06:21,360
And the probabilities of these
events are the probabilities

109
00:06:21,360 --> 00:06:25,860
of the different values of the
random variable, X. So the

110
00:06:25,860 --> 00:06:27,790
probabilities of these
different values

111
00:06:27,790 --> 00:06:31,650
should also add to 1.

112
00:06:31,650 --> 00:06:34,950
Let us now go through a simple
example to illustrate the

113
00:06:34,950 --> 00:06:37,950
general method for calculating
the PMF of a

114
00:06:37,950 --> 00:06:39,890
discrete random variable.

115
00:06:39,890 --> 00:06:43,090
We will revisit our familiar
example involving two rolls of

116
00:06:43,090 --> 00:06:44,690
the tetrahedral die.

117
00:06:44,690 --> 00:06:49,000
And let X be the result of the
first roll, Y be the result of

118
00:06:49,000 --> 00:06:50,159
the second roll.

119
00:06:50,159 --> 00:06:53,430
And notice that we're using
uppercase letters.

120
00:06:53,430 --> 00:06:58,180
And this is because X and
Y are random variables.

121
00:06:58,180 --> 00:07:01,160
In order to do any probability
calculations, we also need the

122
00:07:01,160 --> 00:07:02,170
probability law.

123
00:07:02,170 --> 00:07:05,150
So to keep things simple, let us
assume that every possible

124
00:07:05,150 --> 00:07:08,440
outcome, there's 16 of them, has
the same probability which

125
00:07:08,440 --> 00:07:12,420
is therefore 1 over 16 for
each one of the outcomes.

126
00:07:12,420 --> 00:07:15,770
We will concentrate on a
particular random variable

127
00:07:15,770 --> 00:07:20,540
defined to be the sum of the
random variables, X and Y. So

128
00:07:20,540 --> 00:07:24,240
if X and Y both happen to
be 1, then Z will take

129
00:07:24,240 --> 00:07:25,850
the value of 2.

130
00:07:25,850 --> 00:07:29,850
If X is 2 and Y is 1 our random
variable will take the

131
00:07:29,850 --> 00:07:30,940
value of 3.

132
00:07:30,940 --> 00:07:35,290
And similarly if we have this
outcome, in those outcomes

133
00:07:35,290 --> 00:07:38,540
here, the random variable
takes the value of 4.

134
00:07:38,540 --> 00:07:43,560
And we can continue this way
by marking, for each

135
00:07:43,560 --> 00:07:47,040
particular outcome, the
corresponding value of the

136
00:07:47,040 --> 00:07:48,735
random variable of interest.

137
00:07:52,030 --> 00:07:56,280
What we want to do now is to
calculate the PMF of this

138
00:07:56,280 --> 00:07:57,110
random variable.

139
00:07:57,110 --> 00:07:59,690
What does it mean to
calculate the PMF?

140
00:07:59,690 --> 00:08:07,600
We need to find this value for
all choices of z, that is for

141
00:08:07,600 --> 00:08:11,530
all possible values in the range
of our random variable.

142
00:08:11,530 --> 00:08:15,120
The way we're going to do it is
to consider each possible

143
00:08:15,120 --> 00:08:19,830
value of z, one at a time, and
for any particular value find

144
00:08:19,830 --> 00:08:21,810
out what are the outcomes--

145
00:08:21,810 --> 00:08:24,110
the elements of the
sample space--

146
00:08:24,110 --> 00:08:28,220
for which our random variable
takes on the specific value,

147
00:08:28,220 --> 00:08:31,240
and add the probabilities
of those outcomes.

148
00:08:31,240 --> 00:08:35,820
So to illustrate this process,
let us calculate the value of

149
00:08:35,820 --> 00:08:39,630
the PMF for z equal to 2.

150
00:08:39,630 --> 00:08:42,900
This is by definition the
probability that our random

151
00:08:42,900 --> 00:08:45,950
variable takes the value of 2.

152
00:08:45,950 --> 00:08:49,640
And this is an event that
can only happen here.

153
00:08:49,640 --> 00:08:52,690
It corresponds to only one
element of the sample space,

154
00:08:52,690 --> 00:08:55,185
which has probability
1 over 16.

155
00:08:57,970 --> 00:09:02,390
We can continue the same way
for other values of z.

156
00:09:02,390 --> 00:09:07,930
So for example, the value of PMF
at z equal to 3, this is

157
00:09:07,930 --> 00:09:10,640
the probability that our random
variable takes the

158
00:09:10,640 --> 00:09:12,150
value of 3.

159
00:09:12,150 --> 00:09:15,300
This is an event that can
happen in two ways--

160
00:09:15,300 --> 00:09:17,810
it corresponds to
two outcomes--

161
00:09:17,810 --> 00:09:22,540
and so it has probability
2 over 16.

162
00:09:22,540 --> 00:09:26,100
Continuing similarly, the
probability that our random

163
00:09:26,100 --> 00:09:33,650
variable takes the value of
4 is equal to 3 over 16.

164
00:09:33,650 --> 00:09:37,250
And we can continue this way
and calculate the remaining

165
00:09:37,250 --> 00:09:39,080
entries of our PMF.

166
00:09:39,080 --> 00:09:42,330
After you are done, you
end up with a table--

167
00:09:42,330 --> 00:09:44,850
or rather a graph--

168
00:09:44,850 --> 00:09:47,050
a plot that has this form.

169
00:09:47,050 --> 00:09:50,590
And these are the values of the
different probabilities

170
00:09:50,590 --> 00:09:52,005
that we have computed.

171
00:09:54,630 --> 00:09:57,260
And you can continue with
the other values.

172
00:09:57,260 --> 00:10:00,030
It's a reasonable guess that
this was going to be 4 over

173
00:10:00,030 --> 00:10:07,490
16, this is going to be 3 over
16, 2 over 16, and 1 over 16.

174
00:10:07,490 --> 00:10:10,640
So we have completely determined
the PMF of our

175
00:10:10,640 --> 00:10:11,640
random variable.

176
00:10:11,640 --> 00:10:14,950
We have given the form
of the answers.

177
00:10:14,950 --> 00:10:18,790
And it's always convenient to
also provide a plot with the

178
00:10:18,790 --> 00:10:20,040
answers that we have.