WEBVTT

00:00:01.340 --> 00:00:04.440
In this segment and the next
two, we will introduce a few

00:00:04.440 --> 00:00:08.590
useful random variables that
show up in many applications--

00:00:08.590 --> 00:00:11.730
discrete uniform random
variables, binomial random

00:00:11.730 --> 00:00:14.870
variables, and geometric random
variables So let's

00:00:14.870 --> 00:00:17.030
start with a discrete uniform.

00:00:17.030 --> 00:00:21.460
A discrete uniform random
variable is one that has a PMF

00:00:21.460 --> 00:00:23.060
of this form.

00:00:23.060 --> 00:00:27.530
It takes values in a certain
range, and each one of the

00:00:27.530 --> 00:00:31.060
values in that range has
the same probability.

00:00:31.060 --> 00:00:35.910
To be more precise, a discrete
uniform is completely

00:00:35.910 --> 00:00:39.970
determined by two parameters
that are two integers, a and

00:00:39.970 --> 00:00:44.290
b, which are the beginning and
the end of the range of that

00:00:44.290 --> 00:00:45.790
random variable.

00:00:45.790 --> 00:00:47.930
We're thinking of an experiment
where we're going

00:00:47.930 --> 00:00:54.570
to pick an integer at random
among the values that are

00:00:54.570 --> 00:00:59.980
between a and b with the end
points a and b included.

00:00:59.980 --> 00:01:03.120
And all of these values
are equally likely.

00:01:03.120 --> 00:01:07.230
To be more formal, our sample
space is the set of integers

00:01:07.230 --> 00:01:09.402
from a until b.

00:01:09.402 --> 00:01:14.890
And the number of points that we
have in our sample space is

00:01:14.890 --> 00:01:19.730
b minus a plus 1 possible
values.

00:01:25.120 --> 00:01:27.780
What is the random variable
that we're talking about?

00:01:27.780 --> 00:01:31.720
If this is our sample space, the
outcome of the experiment

00:01:31.720 --> 00:01:33.410
is already a number.

00:01:33.410 --> 00:01:37.210
And the numerical value of the
random variable is just the

00:01:37.210 --> 00:01:40.820
number that we happen to
pick in that range.

00:01:40.820 --> 00:01:43.750
So in this context, there isn't
really a distinction

00:01:43.750 --> 00:01:46.650
between the outcome of the
experiment and the numerical

00:01:46.650 --> 00:01:48.390
value of the random variable.

00:01:48.390 --> 00:01:51.310
They are one in the same.

00:01:51.310 --> 00:01:55.090
Now since each one of the values
is equally likely, and

00:01:55.090 --> 00:01:58.210
since we have so many possible
values, this means that the

00:01:58.210 --> 00:02:04.040
probability of any particular
value is going to be 1 over b

00:02:04.040 --> 00:02:06.400
minus a plus 1.

00:02:06.400 --> 00:02:09.330
This is the choice for the
probability that would make

00:02:09.330 --> 00:02:14.270
all the probabilities in
the PMF sum to one.

00:02:14.270 --> 00:02:18.160
What does this random variable
model in the real world?

00:02:18.160 --> 00:02:22.440
It models a case where we have
a range of possible values,

00:02:22.440 --> 00:02:25.000
and we have complete ignorance,
no reason to

00:02:25.000 --> 00:02:28.620
believe that one value is more
likely than the other.

00:02:28.620 --> 00:02:31.610
As an example, suppose that
you look at your digital

00:02:31.610 --> 00:02:34.050
clock, and you look
at the time.

00:02:34.050 --> 00:02:41.780
And the time that it tells you
is 11:52 and 26 seconds.

00:02:41.780 --> 00:02:45.960
And suppose that you just
look at the seconds.

00:02:45.960 --> 00:02:50.329
The seconds reading is something
that takes values in

00:02:50.329 --> 00:02:53.520
the set from 0 to 59.

00:02:53.520 --> 00:02:56.030
So there are 60 possible
values.

00:02:56.030 --> 00:03:00.440
And if you just choose to look
at your clock at a completely

00:03:00.440 --> 00:03:03.950
random time, there's no reason
to expect that one reading

00:03:03.950 --> 00:03:05.940
would be more likely
than the other.

00:03:05.940 --> 00:03:08.980
All readings should be equally
likely, and each one of them

00:03:08.980 --> 00:03:13.728
should have a probability
of 1 over 60.

00:03:13.728 --> 00:03:17.250
One final comment--

00:03:17.250 --> 00:03:21.650
let us look at the special case
where the beginning and

00:03:21.650 --> 00:03:25.230
the endpoint of the range of
possible values is the same,

00:03:25.230 --> 00:03:28.465
which means that our random
variable can only take one

00:03:28.465 --> 00:03:32.570
value, namely that particular
number a.

00:03:32.570 --> 00:03:35.350
In that case, the random
variable that we're dealing

00:03:35.350 --> 00:03:37.579
with is really a constant.

00:03:37.579 --> 00:03:38.750
It doesn't have any
randomness.

00:03:38.750 --> 00:03:42.440
It is a deterministic random
variable that takes a

00:03:42.440 --> 00:03:47.700
particular value of a with
probability equal to 1.

00:03:47.700 --> 00:03:51.890
It is not random in the common
sense of the world, but

00:03:51.890 --> 00:03:56.690
mathematically we can still
consider it a random variable

00:03:56.690 --> 00:04:00.520
that just happens to be the
same no matter what the

00:04:00.520 --> 00:04:02.240
outcome of the experiment is.