WEBVTT
00:00:01.140 --> 00:00:02.290
Hey guys.
00:00:02.290 --> 00:00:03.570
Welcome back.
00:00:03.570 --> 00:00:07.430
Today, we're going to be working
on a problem that asks
00:00:07.430 --> 00:00:12.860
you to find the PMF of a
function of a random variable.
00:00:12.860 --> 00:00:16.410
So let's just jump right in.
00:00:16.410 --> 00:00:21.060
The problem statement gives
you the PMF for a random
00:00:21.060 --> 00:00:23.180
variable called x.
00:00:23.180 --> 00:00:26.340
So we're told that there's this
random variable x that
00:00:26.340 --> 00:00:32.100
takes on values minus 3, minus
2, minus 1, 1, 2, and 3.
00:00:32.100 --> 00:00:35.440
And for each of those values,
the probability mass lying
00:00:35.440 --> 00:00:39.030
over that value is given
by this formula, x
00:00:39.030 --> 00:00:41.010
squared over a.
00:00:41.010 --> 00:00:44.360
Now I didn't write it here to
save room, but we're also told
00:00:44.360 --> 00:00:47.980
that a is a real number that
is greater than 0.
00:00:47.980 --> 00:00:52.160
And we're told that the
probability of x taking on any
00:00:52.160 --> 00:00:56.430
value outside of the set is 0.
00:00:56.430 --> 00:00:58.930
Now we're asked to do two
things in the problem.
00:00:58.930 --> 00:01:03.040
First is to find the value
of the parameter a.
00:01:03.040 --> 00:01:06.350
And that's sort of a natural
question to ask, because if
00:01:06.350 --> 00:01:11.260
you think about it, the PMF
isn't fully specified.
00:01:11.260 --> 00:01:16.000
And in fact, if you plug in the
wrong number for a, you
00:01:16.000 --> 00:01:18.800
actually won't get
a valid PMF.
00:01:18.800 --> 00:01:22.330
So we'll explore that idea
in the first part.
00:01:22.330 --> 00:01:24.600
And then the second part, you're
given a new random
00:01:24.600 --> 00:01:25.905
variable called z.
00:01:25.905 --> 00:01:28.550
And z happens to be
a function of x.
00:01:28.550 --> 00:01:30.990
In fact, it's equal
to x squared.
00:01:30.990 --> 00:01:34.370
And then you're asked
to compute that PMF.
00:01:34.370 --> 00:01:38.810
So this problem is a good
practice problem.
00:01:38.810 --> 00:01:41.730
I think, at this point, you
guys are sort of newly
00:01:41.730 --> 00:01:45.360
acquainted with the
idea of a PMF, or
00:01:45.360 --> 00:01:47.340
probability mass function.
00:01:47.340 --> 00:01:50.110
So this problem will hopefully
help you get more familiar
00:01:50.110 --> 00:01:55.570
with that concept and how
to manipulate PMFs.
00:01:55.570 --> 00:01:58.060
And by the way, just to make
sure we're all on the same
00:01:58.060 --> 00:02:02.360
page, what does a PMF
really tell you?
00:02:02.360 --> 00:02:08.389
So p sub X, where this is
a capital X, because the
00:02:08.389 --> 00:02:10.520
convention in this class
is to use capital
00:02:10.520 --> 00:02:12.350
letters for random variables.
00:02:12.350 --> 00:02:17.700
So p X of k, this is defined
to be the probability that
00:02:17.700 --> 00:02:22.740
your random variable X takes
on a value of k.
00:02:22.740 --> 00:02:25.630
So essentially, this says--
00:02:25.630 --> 00:02:27.750
and this is just some number.
00:02:27.750 --> 00:02:31.290
So in our particular case,
this would be equal to k
00:02:31.290 --> 00:02:33.260
squared over a.
00:02:33.260 --> 00:02:38.470
And how you can interpret
this is this px guy is
00:02:38.470 --> 00:02:40.590
sort of like a machine.
00:02:40.590 --> 00:02:45.950
He takes in some value that
your random variable could
00:02:45.950 --> 00:02:50.340
take on, and then he spits out
the amount of probability mass
00:02:50.340 --> 00:02:53.010
lying over that value.
00:02:53.010 --> 00:02:53.330
OK.
00:02:53.330 --> 00:02:57.930
So now that we've done that
quick recap, let's get back to
00:02:57.930 --> 00:02:59.530
the first part of the problem.
00:02:59.530 --> 00:03:03.530
So we have this formula for
px of x, and we need
00:03:03.530 --> 00:03:04.930
to solve for a.
00:03:04.930 --> 00:03:08.990
So in order to do that, we're
going to use one of our axioms
00:03:08.990 --> 00:03:11.160
of probability to set
up an equation.
00:03:11.160 --> 00:03:13.860
And then we can solve
precisely for a.
00:03:13.860 --> 00:03:20.120
So namely, we know that every
PMF must sum to 1.
00:03:20.120 --> 00:03:24.000
And so essentially, if you sum
this guy over all possible
00:03:24.000 --> 00:03:27.410
values of x, you should get a 1,
and that equation will let
00:03:27.410 --> 00:03:29.060
us solve for a.
00:03:29.060 --> 00:03:30.910
So let's do that.
00:03:30.910 --> 00:03:37.090
Summation over x of px of x.
00:03:37.090 --> 00:03:39.430
So here, essentially
you're only summing
00:03:39.430 --> 00:03:42.230
over these six values.
00:03:42.230 --> 00:03:49.230
So this is equal to px of minus
3, plus px of minus 2,
00:03:49.230 --> 00:03:52.872
plus px of minus 1, et cetera.
00:03:52.872 --> 00:03:58.270
Oops. px of 2 plus px of 3.
00:03:58.270 --> 00:03:58.760
OK.
00:03:58.760 --> 00:04:01.510
And again, like the
interpretation as we said,
00:04:01.510 --> 00:04:08.460
this number here should be
interpreted as the amount of
00:04:08.460 --> 00:04:13.810
probability mass lying
over minus 3.
00:04:13.810 --> 00:04:16.790
And to help you visualize this,
actually, before we go
00:04:16.790 --> 00:04:21.870
further with the computation,
let's actually plot this PMF.
00:04:21.870 --> 00:04:27.140
So the amount of probability
mass lying over minus 3, the
00:04:27.140 --> 00:04:31.000
way we figure that out is we
take minus 3 and we plug it
00:04:31.000 --> 00:04:32.780
into this formula up here.
00:04:32.780 --> 00:04:34.030
So you get 9/a.
00:04:38.480 --> 00:04:40.750
Now you can do this
for minus 2.
00:04:40.750 --> 00:04:43.740
You've got 4/a, looking
at the formula.
00:04:43.740 --> 00:04:46.140
For 1, you get 1/a.
00:04:46.140 --> 00:04:52.130
And of course, this graph, it's
the mirror image over 0,
00:04:52.130 --> 00:04:55.370
because of the symmetry.
00:04:55.370 --> 00:05:00.600
So hopefully this little
visualization helps you
00:05:00.600 --> 00:05:01.920
understand what I'm
talking about.
00:05:01.920 --> 00:05:04.350
And now we can just read
these values off of the
00:05:04.350 --> 00:05:06.470
plot we just made.
00:05:06.470 --> 00:05:10.310
So we know px minus 3
is equal to px of 3.
00:05:10.310 --> 00:05:15.330
So we can go ahead and just
take 2 times 9/a.
00:05:15.330 --> 00:05:22.700
Similarly, we get 2 times 4/a,
and then plus 2 times 1/a.
00:05:22.700 --> 00:05:25.500
So now it's just a question
of algebra.
00:05:25.500 --> 00:05:30.670
So simplifying this, you're
going to get 18 plus 8 plus 2,
00:05:30.670 --> 00:05:32.860
divided by a.
00:05:32.860 --> 00:05:35.750
And this gives you 28/a.
00:05:35.750 --> 00:05:40.180
And as I argued before, you
know that if you sum a PMF
00:05:40.180 --> 00:05:43.120
over all possible values,
you must get 1.
00:05:43.120 --> 00:05:46.910
So this is equal to 1, which
of course implies that a is
00:05:46.910 --> 00:05:48.160
equal to 28.
00:05:50.330 --> 00:05:54.760
So what we've shown here is that
you actually don't have a
00:05:54.760 --> 00:05:59.360
choice for what value
a can take on.
00:05:59.360 --> 00:06:01.790
It must take on 28.
00:06:01.790 --> 00:06:05.850
And in fact, if you plug in
any other value than 28 in
00:06:05.850 --> 00:06:08.510
here, you actually are not going
to have a valid PMF,
00:06:08.510 --> 00:06:11.810
because it's not going
to sum to 1.
00:06:11.810 --> 00:06:12.250
OK.
00:06:12.250 --> 00:06:16.960
So I'm going to write my answer
here, and then erase to
00:06:16.960 --> 00:06:20.580
give myself more room for part