WEBVTT
00:00:00.230 --> 00:00:03.560
In this lecture we complete
our discussion of multiple
00:00:03.560 --> 00:00:06.350
continuous random variables.
00:00:06.350 --> 00:00:08.850
In the first half, we talk
about the conditional
00:00:08.850 --> 00:00:11.160
distribution of one random
variable, given
00:00:11.160 --> 00:00:13.160
the value of another.
00:00:13.160 --> 00:00:16.120
We will see that the mechanics
are essentially the same as in
00:00:16.120 --> 00:00:18.180
the discrete case.
00:00:18.180 --> 00:00:21.950
Here, we will actually face some
subtle issues, because we
00:00:21.950 --> 00:00:26.040
will be conditioning on any
event that has 0 probability.
00:00:26.040 --> 00:00:29.590
Nevertheless, all formulas will
still have the form that
00:00:29.590 --> 00:00:30.960
one should expect.
00:00:30.960 --> 00:00:34.220
And in particular, we will see
natural versions of the total
00:00:34.220 --> 00:00:38.580
probability and total
expectation theorems.
00:00:38.580 --> 00:00:41.510
We will also define independence
of continuous
00:00:41.510 --> 00:00:44.850
random variables, a concept that
has the same intuitive
00:00:44.850 --> 00:00:47.290
content as in the
discrete case.
00:00:47.290 --> 00:00:50.410
That is, when we have
independent random variables,
00:00:50.410 --> 00:00:54.420
the values of some of them do
not cause any revision of our
00:00:54.420 --> 00:00:57.590
beliefs about the
remaining ones.
00:00:57.590 --> 00:01:01.090
Then, in the second half of the
lecture, we will focus on
00:01:01.090 --> 00:01:02.550
the Bayes rule.
00:01:02.550 --> 00:01:05.069
This will be the methodological
foundation for
00:01:05.069 --> 00:01:08.120
when, later in this course,
we dive into
00:01:08.120 --> 00:01:10.490
the subject of inference.
00:01:10.490 --> 00:01:13.660
The Bayes rule allows us to
revise our beliefs about a
00:01:13.660 --> 00:01:14.670
random variable.
00:01:14.670 --> 00:01:18.370
That is, replace an original
probability distribution by a
00:01:18.370 --> 00:01:21.920
conditional one, after we
observe the value of some
00:01:21.920 --> 00:01:24.770
other random variable.
00:01:24.770 --> 00:01:27.539
Depending on whether the random
variables involved are
00:01:27.539 --> 00:01:31.660
discrete or continuous, we
will get four different
00:01:31.660 --> 00:01:33.890
versions of the Bayes rule.
00:01:33.890 --> 00:01:37.970
They all have the same form,
with small differences.
00:01:37.970 --> 00:01:40.740
And we will see how to apply
them through some examples.