1
00:00:00,000 --> 00:00:04,320
The coefficients n-choose-k
that we calculated in the

2
00:00:04,320 --> 00:00:08,220
previous segment are known as
the binomial coefficients.

3
00:00:08,220 --> 00:00:11,060
They are intimately related
to certain probabilities

4
00:00:11,060 --> 00:00:15,590
associated with coin tossing
models, the so-called binomial

5
00:00:15,590 --> 00:00:16,930
probabilities.

6
00:00:16,930 --> 00:00:19,670
This is going to
be our subject.

7
00:00:19,670 --> 00:00:25,540
We consider a coin which we
toss n times in a row,

8
00:00:25,540 --> 00:00:28,190
independently.

9
00:00:28,190 --> 00:00:31,840
For each one of the tosses of
this coin, we assume that

10
00:00:31,840 --> 00:00:34,360
there is a certain probability,
p, that the

11
00:00:34,360 --> 00:00:38,290
result is heads, which of
course, implies that the

12
00:00:38,290 --> 00:00:42,340
probability of obtaining tails
in any particular toss is

13
00:00:42,340 --> 00:00:45,180
going to be 1 minus p.

14
00:00:45,180 --> 00:00:49,740
The question we want to address
is the following.

15
00:00:49,740 --> 00:00:53,400
We want to calculate the
probability that in those n

16
00:00:53,400 --> 00:00:56,010
independent coin tosses,
we're going to

17
00:00:56,010 --> 00:00:58,835
observe exactly k heads.

18
00:01:02,060 --> 00:01:05,980
Let us start working our way
towards the solution to this

19
00:01:05,980 --> 00:01:10,600
problem by looking first
at a simple setting

20
00:01:10,600 --> 00:01:13,300
and then move on.

21
00:01:13,300 --> 00:01:16,600
So let us answer this
simple question.

22
00:01:16,600 --> 00:01:19,530
What is the probability
that we observe

23
00:01:19,530 --> 00:01:21,940
this particular sequence?

24
00:01:21,940 --> 00:01:26,780
Of course here we take n equal
to six, and we wish to

25
00:01:26,780 --> 00:01:29,020
calculate this probability.

26
00:01:29,020 --> 00:01:32,200
Now, because we have assumed
that the coin tosses are

27
00:01:32,200 --> 00:01:35,250
independent, we can multiply
probabilities.

28
00:01:35,250 --> 00:01:38,000
So the probability of this
sequence is equal to the

29
00:01:38,000 --> 00:01:42,690
probability that the first
toss is heads times the

30
00:01:42,690 --> 00:01:45,850
probability that the second
toss is tails, which is 1

31
00:01:45,850 --> 00:01:49,860
minus p, times the probability
that the third toss is tails,

32
00:01:49,860 --> 00:01:53,460
which is 1 minus p, times the
probability of heads, times

33
00:01:53,460 --> 00:01:54,820
the probability of
heads, times the

34
00:01:54,820 --> 00:01:56,289
probability of heads.

35
00:01:56,289 --> 00:02:01,290
And by collecting terms, this is
p to the 4th times 1 minus

36
00:02:01,290 --> 00:02:02,620
p to the second power.

37
00:02:05,880 --> 00:02:11,840
More generally, if I give you a
particular sequence of heads

38
00:02:11,840 --> 00:02:15,910
and tails, as in this example,
and I ask you, what is the

39
00:02:15,910 --> 00:02:18,710
probability that this particular
sequence is

40
00:02:18,710 --> 00:02:25,160
observed, then by generalizing
from this answer or from the

41
00:02:25,160 --> 00:02:31,170
derivation of this answer, you
see that you're going to get p

42
00:02:31,170 --> 00:02:33,615
to the power number of heads.

43
00:02:36,120 --> 00:02:39,940
And the reason is that each
time that there's a head

44
00:02:39,940 --> 00:02:43,220
showing up in this sequence,
there's a corresponding factor

45
00:02:43,220 --> 00:02:46,570
of p in this numerical answer.

46
00:02:46,570 --> 00:02:50,450
And then there are factors
associated with tails.

47
00:02:50,450 --> 00:02:54,050
Each tail contributes a
factor of 1 minus p.

48
00:02:54,050 --> 00:02:57,980
And so we're going to have
here 1 minus p to a power

49
00:02:57,980 --> 00:02:59,910
equal to the number of tails.

50
00:03:04,050 --> 00:03:08,270
Now, if I ask you about the
probability of a particular

51
00:03:08,270 --> 00:03:11,720
sequence and that particular
sequence has happened to have

52
00:03:11,720 --> 00:03:15,400
exactly k heads, what is the
probability of that sequence?

53
00:03:15,400 --> 00:03:18,340
Well, we already calculated
what it is.

54
00:03:18,340 --> 00:03:22,920
It is the previous answer,
except we use the symbol k

55
00:03:22,920 --> 00:03:27,560
instead of just writing out
explicitly "number of heads."

56
00:03:27,560 --> 00:03:31,960
And the number of tails is the
number of tosses minus how

57
00:03:31,960 --> 00:03:36,400
many tosses resulted in heads.

58
00:03:36,400 --> 00:03:40,810
Now, we're ready to consider
the actual problem that we

59
00:03:40,810 --> 00:03:43,100
want to solve, which
is calculate the

60
00:03:43,100 --> 00:03:46,010
probability of k heads.

61
00:03:46,010 --> 00:03:50,270
The event of obtaining
k heads can happen in

62
00:03:50,270 --> 00:03:52,490
many different ways.

63
00:03:52,490 --> 00:03:58,280
Any particular k-head sequence
makes that event to occur.

64
00:03:58,280 --> 00:04:04,590
Any particular k-head sequence
has a probability equal to

65
00:04:04,590 --> 00:04:06,840
this expression.

66
00:04:06,840 --> 00:04:11,170
The overall probability of k
heads is going to be the

67
00:04:11,170 --> 00:04:18,230
probability of any particular
k-head sequence, times the

68
00:04:18,230 --> 00:04:22,170
number of k-head sequences
that we have.

69
00:04:27,490 --> 00:04:31,800
Now, the reason why we can carry
out this argument is the

70
00:04:31,800 --> 00:04:36,090
fact that any k-head sequence
has the same probability.

71
00:04:36,090 --> 00:04:39,520
Otherwise, we wouldn't be able
to write down an answer which

72
00:04:39,520 --> 00:04:41,980
is just the product
of two terms.

73
00:04:41,980 --> 00:04:44,600
But because every k-head
sequence has the same

74
00:04:44,600 --> 00:04:47,860
probability, to find the overall
probability, we take

75
00:04:47,860 --> 00:04:52,050
the probability of each one of
them and multiply it with the

76
00:04:52,050 --> 00:04:55,360
number of how many
of these we have.

77
00:04:55,360 --> 00:05:00,110
So to make further progress, now
we need to calculate the

78
00:05:00,110 --> 00:05:03,550
number of possible
k-head sequences.

79
00:05:03,550 --> 00:05:05,920
How many are there?

80
00:05:05,920 --> 00:05:09,340
Well, specifying a k-head
sequence is

81
00:05:09,340 --> 00:05:11,160
the same as the following.

82
00:05:11,160 --> 00:05:14,580
You think of having
n time slots.

83
00:05:14,580 --> 00:05:17,590
These time slots corresponds
to the different

84
00:05:17,590 --> 00:05:20,240
tosses of your coin.

85
00:05:20,240 --> 00:05:24,410
And to specify a k-head
sequence, you need to tell me

86
00:05:24,410 --> 00:05:31,610
which ones of these slots happen
to contain a head.

87
00:05:31,610 --> 00:05:35,420
You need to tell me
k of those slots.

88
00:05:35,420 --> 00:05:40,800
So in other words, what you're
doing is you're specifying a

89
00:05:40,800 --> 00:05:45,460
subset of the set of
these n slots, a

90
00:05:45,460 --> 00:05:48,290
subset that has k elements.

91
00:05:48,290 --> 00:05:53,070
You need to choose k of the
slots out of the n and tell me

92
00:05:53,070 --> 00:05:55,630
that those k slots have heads.

93
00:05:55,630 --> 00:05:59,560
That's the way of specifying a
particular k-head sequence.

94
00:05:59,560 --> 00:06:02,630
So what's the number of
k-head sequences?

95
00:06:02,630 --> 00:06:05,760
Well, it's the same as the
number of ways that you can

96
00:06:05,760 --> 00:06:11,110
choose k slots out of the n
slots, which is our binomial

97
00:06:11,110 --> 00:06:14,410
coefficient, n-choose-k.

98
00:06:14,410 --> 00:06:18,420
Therefore, the answer to our
problem is this expression

99
00:06:18,420 --> 00:06:25,590
here, times n-choose-k, which
is shown up here.

100
00:06:25,590 --> 00:06:30,080
At this point, we can pause and
consider a simple question

101
00:06:30,080 --> 00:06:34,040
to check your understanding of
the binomial probabilities.