1
00:00:00,000 --> 00:00:00,550

2
00:00:00,550 --> 00:00:02,250
Welcome back guys.

3
00:00:02,250 --> 00:00:05,080
Today we're going to work on
a problem that tests your

4
00:00:05,080 --> 00:00:07,550
knowledge of joint PMFs.

5
00:00:07,550 --> 00:00:09,620
And we're also going to get
some practice computing

6
00:00:09,620 --> 00:00:13,200
conditional expectations and
conditional variances.

7
00:00:13,200 --> 00:00:15,740
So in this problem, we
are given a set of

8
00:00:15,740 --> 00:00:17,610
points in the xy plane.

9
00:00:17,610 --> 00:00:20,290
And we're told that these points
are equally likely.

10
00:00:20,290 --> 00:00:21,210
So there's eight of them.

11
00:00:21,210 --> 00:00:22,410
And each point has a

12
00:00:22,410 --> 00:00:25,500
probability of 1/8 of occurring.

13
00:00:25,500 --> 00:00:27,900
And we're also given this
list of questions.

14
00:00:27,900 --> 00:00:30,670
And we're going to work
through them together.

15
00:00:30,670 --> 00:00:34,590
So in part a, we are asked to
find the values of x that

16
00:00:34,590 --> 00:00:39,010
maximize the conditional
expectation of y given x.

17
00:00:39,010 --> 00:00:42,300
So jumping right in, this
is the quantity

18
00:00:42,300 --> 00:00:45,190
we're interested in.

19
00:00:45,190 --> 00:00:50,610
And so this quantity
is a function of x.

20
00:00:50,610 --> 00:00:53,430
You plug-in various
values of x.

21
00:00:53,430 --> 00:00:56,630
And then this will spit
out a scalar value.

22
00:00:56,630 --> 00:00:59,970
And that value will correspond
to the conditional expectation

23
00:00:59,970 --> 00:01:04,640
of y conditioned on the value
of x that you put in.

24
00:01:04,640 --> 00:01:09,690
So let's see, when x is equal
to 0, for instance, let's

25
00:01:09,690 --> 00:01:11,410
figure out what this value is.

26
00:01:11,410 --> 00:01:15,790
Well, when x is equal to 0
we're living in a world,

27
00:01:15,790 --> 00:01:17,660
essentially, on this line.

28
00:01:17,660 --> 00:01:19,520
So that means that
only these two

29
00:01:19,520 --> 00:01:20,730
points could have occurred.

30
00:01:20,730 --> 00:01:25,550
And in particular, y can only
take on the values of 1 and 3.

31
00:01:25,550 --> 00:01:28,580
Now, since all these points in
the unconditional universe

32
00:01:28,580 --> 00:01:31,740
were equally likely, in the
conditional universe they will

33
00:01:31,740 --> 00:01:33,530
still be equally likely.

34
00:01:33,530 --> 00:01:35,620
So this happens with
probability 1/2.

35
00:01:35,620 --> 00:01:38,640
And this happens with
probability 1/2.

36
00:01:38,640 --> 00:01:45,960
And therefore, the expectation
would just be 3/2 plus 1/2

37
00:01:45,960 --> 00:01:49,380
which is 4/2, or 2.

38
00:01:49,380 --> 00:01:51,700
But a much faster way of seeing
this-- and it's the

39
00:01:51,700 --> 00:01:54,300
strategy that I'm going to use
for the rest of the problem--

40
00:01:54,300 --> 00:01:56,720
is to remember that
expectation acts

41
00:01:56,720 --> 00:01:58,780
like center of mass.

42
00:01:58,780 --> 00:02:01,490
So the center of mass, when
these two points are equally

43
00:02:01,490 --> 00:02:05,010
likely, is just the midpoint,
which of course is 2.

44
00:02:05,010 --> 00:02:09,400
So we're going to use that
intuition on the other ones.

45
00:02:09,400 --> 00:02:12,730
So I'm skipping to x is
equal to 2 because 1

46
00:02:12,730 --> 00:02:14,230
and 3 are not possible.

47
00:02:14,230 --> 00:02:17,000
So when x is equal to 2,
y can only take on the

48
00:02:17,000 --> 00:02:18,490
values of 1 or 2.

49
00:02:18,490 --> 00:02:20,240
Again, they're equally likely.

50
00:02:20,240 --> 00:02:22,390
So the center of mass is in the
middle which happens at

51
00:02:22,390 --> 00:02:24,000
1.5 or 3/2.

52
00:02:24,000 --> 00:02:26,540

53
00:02:26,540 --> 00:02:29,230
Similarly, x is equal to 4.

54
00:02:29,230 --> 00:02:31,736
We're living in this conditional
universe, where y

55
00:02:31,736 --> 00:02:34,380
can take on of these
four points with

56
00:02:34,380 --> 00:02:36,870
probability 1/4 each.

57
00:02:36,870 --> 00:02:39,920
And so again, we expect
the center of mass to

58
00:02:39,920 --> 00:02:42,850
be at 1.5 or 3/2.

59
00:02:42,850 --> 00:02:45,555
And this quantity is undefined
otherwise.

60
00:02:45,555 --> 00:02:48,650

61
00:02:48,650 --> 00:02:50,540
OK, so we're almost done.

62
00:02:50,540 --> 00:02:55,240
Now we just need to find which
value of x maximizes this.

63
00:02:55,240 --> 00:02:58,460
Well, let's see, 2 is the
biggest quantity out of all of

64
00:02:58,460 --> 00:02:59,630
these numbers.

65
00:02:59,630 --> 00:03:01,320
So the maximum is 2.

66
00:03:01,320 --> 00:03:04,090
And it occurs when
x is equal to 0.

67
00:03:04,090 --> 00:03:05,860
So we come over here.

68
00:03:05,860 --> 00:03:07,400
And we found our answer.

69
00:03:07,400 --> 00:03:10,780
x is equal to 0 is the value,
which maximizes the

70
00:03:10,780 --> 00:03:13,930
conditional expectation
of y given x.

71
00:03:13,930 --> 00:03:15,970
So part b is very similar
to part a.

72
00:03:15,970 --> 00:03:18,310
But there is slightly more
computation involved.

73
00:03:18,310 --> 00:03:21,170
Because now we're dealing with
the variance and not an

74
00:03:21,170 --> 00:03:22,395
expectation.

75
00:03:22,395 --> 00:03:28,220
And variance is usually a little
bit tougher to compute.

76
00:03:28,220 --> 00:03:30,250
So we're going to start
in the same manner.

77
00:03:30,250 --> 00:03:34,490
But I want you guys to see if
you can figure out intuitively

78
00:03:34,490 --> 00:03:36,160
what the right value is.

79
00:03:36,160 --> 00:03:38,160
I'm going to do the entire
computation now.

80
00:03:38,160 --> 00:03:41,070
And then you can compare whether
your intuition matches

81
00:03:41,070 --> 00:03:43,320
with the real results.

82
00:03:43,320 --> 00:03:49,560
So variance of x conditioned
on a particular value of y,

83
00:03:49,560 --> 00:03:51,770
this is now a function of y.

84
00:03:51,770 --> 00:03:54,430
For each value of y you plug in
you're going to get out a

85
00:03:54,430 --> 00:03:55,500
scalar number.

86
00:03:55,500 --> 00:03:58,680
And that number represents the
conditional variance of x when

87
00:03:58,680 --> 00:04:02,040
you condition on the value
of y that you plugged in.

88
00:04:02,040 --> 00:04:07,440
So let's see, when y is equal
to 0 we have a nice case.

89
00:04:07,440 --> 00:04:11,680
If y is equal to 0 we have no
freedom about what x is.

90
00:04:11,680 --> 00:04:13,820
This is the only point that
could have occurred.

91
00:04:13,820 --> 00:04:17,279
Therefore, x definitely
takes on a value of 4.

92
00:04:17,279 --> 00:04:19,920
And there's no uncertainty
left.

93
00:04:19,920 --> 00:04:23,220
So in other words, the
variance is 0.

94
00:04:23,220 --> 00:04:30,040
Now, if y is equal to 1, x can
take on a value of 0, a value

95
00:04:30,040 --> 00:04:31,990
of 2 or a value of 4.

96
00:04:31,990 --> 00:04:35,220
And these all have the same
probability of occurring, of

97
00:04:35,220 --> 00:04:36,240
1/3,

98
00:04:36,240 --> 00:04:38,920
And again, the reasoning behind
that is that all eight

99
00:04:38,920 --> 00:04:42,430
points were equally likely in
the unconditional universe.

100
00:04:42,430 --> 00:04:46,990
If you condition on y being
equal to 1 these outcomes

101
00:04:46,990 --> 00:04:48,970
still have the same relative
frequency.

102
00:04:48,970 --> 00:04:51,190
Namely, they're still
equally likely.

103
00:04:51,190 --> 00:04:53,020
And since there are three
of them they now have a

104
00:04:53,020 --> 00:04:55,730
probability of 1/3 each.

105
00:04:55,730 --> 00:05:01,430
So we're going to go ahead and
use a formula that hopefully,

106
00:05:01,430 --> 00:05:04,080
you guys remember.

107
00:05:04,080 --> 00:05:11,010
So in particular, variance is
the expectation of x squared

108
00:05:11,010 --> 00:05:15,520
minus the expectation
of x all squared,

109
00:05:15,520 --> 00:05:16,920
the whole thing squared.

110
00:05:16,920 --> 00:05:20,380
So let's start by computing
this number first.

111
00:05:20,380 --> 00:05:22,770
So conditioned on y
is equal to 1--

112
00:05:22,770 --> 00:05:24,095
so we're in this line--

113
00:05:24,095 --> 00:05:27,140
the expectation of x
is just 2, right?

114
00:05:27,140 --> 00:05:29,660
The same center-of-mass
to argument.

115
00:05:29,660 --> 00:05:34,650
So this, we have a minus
2 squared over here.

116
00:05:34,650 --> 00:05:38,450
Now, x squared is only slightly
more difficult.

117
00:05:38,450 --> 00:05:41,960
With probability 1/3,
x squared will take

118
00:05:41,960 --> 00:05:43,650
on a value of 0.

119
00:05:43,650 --> 00:05:46,510
With probability 1/3,
x squared will take

120
00:05:46,510 --> 00:05:47,990
on a value of 4.

121
00:05:47,990 --> 00:05:49,190
I'm just doing 2 squared.

122
00:05:49,190 --> 00:05:53,180
And with probability 1/3, x
squared takes on a value of 4

123
00:05:53,180 --> 00:05:55,520
squared or 16.

124
00:05:55,520 --> 00:05:58,960
So writing down when I just
said, we have 0 times 1/3

125
00:05:58,960 --> 00:06:00,610
which is 0.

126
00:06:00,610 --> 00:06:04,310
We have 2 squared, which
is 4 times 1/3.

127
00:06:04,310 --> 00:06:08,960
And then we have 4 squared,
which is 16 times 1/3.

128
00:06:08,960 --> 00:06:11,710
And then we have our minus
4 from before.

129
00:06:11,710 --> 00:06:18,260
So doing this math out, we get,
let's see, 20/3 minus

130
00:06:18,260 --> 00:06:23,320
12/3, which is equal
to 8/3, or 8/3.

131
00:06:23,320 --> 00:06:26,790
So we'll come back up
here and put 8/3.

132
00:06:26,790 --> 00:06:30,270
So I realize I'm going through
this pretty quickly.

133
00:06:30,270 --> 00:06:32,280
Hopefully this step didn't
confuse you.

134
00:06:32,280 --> 00:06:35,850
Essentially, when I was doing
is, if you think of x squared

135
00:06:35,850 --> 00:06:40,180
as a new random variable, x
squared, the possible values

136
00:06:40,180 --> 00:06:44,240
that it can take on are 0,
4, and 16 when you're

137
00:06:44,240 --> 00:06:47,380
conditioning on y
is equal to 1.

138
00:06:47,380 --> 00:06:51,490
And so I was simply saying
that that random variable

139
00:06:51,490 --> 00:06:56,960
takes on those values with
equal probability.

140
00:06:56,960 --> 00:06:59,730
So let's move on to
the next one.

141
00:06:59,730 --> 00:07:03,370
So if we condition on y is equal
to 2 we're going to do a

142
00:07:03,370 --> 00:07:06,030
very similar computation.

143
00:07:06,030 --> 00:07:08,180
Oops, I shouldn't have
erased that.

144
00:07:08,180 --> 00:07:10,790
OK, so we're going to use the
same formula that we just

145
00:07:10,790 --> 00:07:16,330
used, which is the expectation
of x given y is equal to 2.

146
00:07:16,330 --> 00:07:21,090
Sorry, x squared minus the
expectation of x conditioned

147
00:07:21,090 --> 00:07:24,040
on y is equal to
2, all squared.

148
00:07:24,040 --> 00:07:26,470
So conditioned on y
is equal to 2, the

149
00:07:26,470 --> 00:07:28,350
expectation of x is 3.

150
00:07:28,350 --> 00:07:30,630
Same center of mass argument.

151
00:07:30,630 --> 00:07:33,150
So 3 squared is 9.

152
00:07:33,150 --> 00:07:38,370
And then x squared can
take on a value of 4.

153
00:07:38,370 --> 00:07:40,870
Or it can take on
a value of 16.

154
00:07:40,870 --> 00:07:43,280
And it does so with
equal probability.

155
00:07:43,280 --> 00:07:51,200
So we get 4/2, 4
plus 16 over 2.

156
00:07:51,200 --> 00:07:55,620
So this is 2 plus 8, which
is 10, minus 9.

157
00:07:55,620 --> 00:07:56,630
That'll give us 1.

158
00:07:56,630 --> 00:08:00,710
So we get a 1 when
y is equal to 2.

159
00:08:00,710 --> 00:08:03,120
And last computation and
then we're done.

160
00:08:03,120 --> 00:08:06,070
I'm still recycling
the same formula.

161
00:08:06,070 --> 00:08:11,000
But now we're conditioning
on y is equal to 3.

162
00:08:11,000 --> 00:08:14,610
And then we'll be done with
this problem, I promise.

163
00:08:14,610 --> 00:08:18,620
OK, so when y is equal to 3 x
can take on the value of 0.

164
00:08:18,620 --> 00:08:20,710
Or it can take on
the value of 4.

165
00:08:20,710 --> 00:08:24,110
Those two points happen with
probability 1/2, 1/2.

166
00:08:24,110 --> 00:08:27,790
So the expectation is right
in the middle which is 2.

167
00:08:27,790 --> 00:08:30,210
So we get a minus 4.

168
00:08:30,210 --> 00:08:34,460
And similarly, x squared can
take on the value of 0.

169
00:08:34,460 --> 00:08:36,830
When x takes on the value of
0-- and that happens with

170
00:08:36,830 --> 00:08:38,530
probability 1/2--

171
00:08:38,530 --> 00:08:41,735
similarly, x squared can take
on the value of 16 when x

172
00:08:41,735 --> 00:08:42,929
takes on the value of 4.

173
00:08:42,929 --> 00:08:45,400
And that happens with
probability 1/2.

174
00:08:45,400 --> 00:08:51,540
So we just have 0/2
plus 16/2 minus 4.

175
00:08:51,540 --> 00:08:55,752
And this gives us 8 minus
4, which is simply 4.

176
00:08:55,752 --> 00:08:59,540
So finally, after all that
computation, we are done.

177
00:08:59,540 --> 00:09:03,010
We have the conditional
variance of x given y.

178
00:09:03,010 --> 00:09:07,070
Again, we're interested in when
this value is largest.

179
00:09:07,070 --> 00:09:11,495
And we see that 4 is the biggest
value in this column.

180
00:09:11,495 --> 00:09:15,840
And this value occurs when
y takes on a value of 3.

181
00:09:15,840 --> 00:09:22,420
So our answer, over here,
is y is equal to 3.

182
00:09:22,420 --> 00:09:25,200
All right, so now we're going to
switch gears in part c and

183
00:09:25,200 --> 00:09:27,680
d a little bit.

184
00:09:27,680 --> 00:09:29,660
And we're going to
be more concerned

185
00:09:29,660 --> 00:09:32,930
with PMFs, et cetera.

186
00:09:32,930 --> 00:09:37,840
So in part c, we're given a
random variable called r which

187
00:09:37,840 --> 00:09:41,250
is defined as the minimum
of x and y.

188
00:09:41,250 --> 00:09:43,970
So for instance, this
is the 0.01.

189
00:09:43,970 --> 00:09:45,710
The minimum of 0 and 1 is 0.

190
00:09:45,710 --> 00:09:50,180
So r would have a
value of 0 here.

191
00:09:50,180 --> 00:09:52,250
Now, we can be a little bit
smarter about this.

192
00:09:52,250 --> 00:09:56,250
If we plot the line,
y is equal to x.

193
00:09:56,250 --> 00:09:57,855
So that looks something
like this.

194
00:09:57,855 --> 00:10:00,360

195
00:10:00,360 --> 00:10:06,780
We see that all of the points
below this line satisfy y

196
00:10:06,780 --> 00:10:08,920
being less or equal to x.

197
00:10:08,920 --> 00:10:12,640
And all the points above this
line have y greater than or

198
00:10:12,640 --> 00:10:14,290
equal to x.

199
00:10:14,290 --> 00:10:18,332
So if y is less than or equal to
x, you hopefully agree that

200
00:10:18,332 --> 00:10:21,660
here the min, or r,
is equal to y.

201
00:10:21,660 --> 00:10:26,510
But over here, the min, r, is
actually equal to x, since x

202
00:10:26,510 --> 00:10:27,440
is always smaller.

203
00:10:27,440 --> 00:10:28,890
So now we can go
ahead quickly.

204
00:10:28,890 --> 00:10:31,760
And I'm going to write the value
of r next each point

205
00:10:31,760 --> 00:10:33,190
using this rule.

206
00:10:33,190 --> 00:10:36,960
So here, r is the value
of y, which is 1.

207
00:10:36,960 --> 00:10:39,270
Here, r is equal to 0.

208
00:10:39,270 --> 00:10:40,410
Here r is 1.

209
00:10:40,410 --> 00:10:42,200
Here r is 2.

210
00:10:42,200 --> 00:10:44,650
Here r is 3.

211
00:10:44,650 --> 00:10:46,600
Over here, r is the
value of x.

212
00:10:46,600 --> 00:10:48,980
So r is equal to 0.

213
00:10:48,980 --> 00:10:50,620
And r is equal to 0 here.

214
00:10:50,620 --> 00:10:53,000
And so the only point we didn't
handle is the one that

215
00:10:53,000 --> 00:10:54,120
lies on the line.

216
00:10:54,120 --> 00:10:55,750
But in that case it's easy.

217
00:10:55,750 --> 00:10:57,250
Because x is equal to 2.

218
00:10:57,250 --> 00:10:59,100
And y is equal to 2.

219
00:10:59,100 --> 00:11:01,590
So the min is simply 2.

220
00:11:01,590 --> 00:11:05,040
So with this information
I claim we're now done.

221
00:11:05,040 --> 00:11:08,690
We can just write down
what the PMF of r is.

222
00:11:08,690 --> 00:11:15,225
So in particular, r takes
on a value of 0.

223
00:11:15,225 --> 00:11:17,820

224
00:11:17,820 --> 00:11:20,200
When this point happens,
this point happens,

225
00:11:20,200 --> 00:11:21,820
or this point happens.

226
00:11:21,820 --> 00:11:22,990
And those collectively have a

227
00:11:22,990 --> 00:11:27,170
probability of 3/8 of occurring.

228
00:11:27,170 --> 00:11:30,420
r can take on a value of 1
when either of these two

229
00:11:30,420 --> 00:11:32,120
points happen.

230
00:11:32,120 --> 00:11:33,830
So that happens with
probability 2/8.

231
00:11:33,830 --> 00:11:36,960

232
00:11:36,960 --> 00:11:38,540
r is equal to 2.

233
00:11:38,540 --> 00:11:40,030
This can happen in two ways.

234
00:11:40,030 --> 00:11:42,450
So we get 2/8.

235
00:11:42,450 --> 00:11:45,740
And r equal to 3 can happen
in only one way.

236
00:11:45,740 --> 00:11:48,170
So we get 1/8.

237
00:11:48,170 --> 00:11:51,710
Quick sanity check, 3 plus
2 is 5, plus 2 is

238
00:11:51,710 --> 00:11:52,920
7, plus 1 is 8.

239
00:11:52,920 --> 00:11:55,230
So our PMF sums to 1.

240
00:11:55,230 --> 00:11:57,120
And to be complete, we
should sketch it.

241
00:11:57,120 --> 00:12:00,220
Because the problem asks
us to sketch it.

242
00:12:00,220 --> 00:12:07,760
So we're plotting PR
of r, 0, 1, 2, 3.

243
00:12:07,760 --> 00:12:11,575
So here we get, let's
see, 1, 2, 3.

244
00:12:11,575 --> 00:12:14,490
For 0 we have 3/8.

245
00:12:14,490 --> 00:12:17,870
For 1 we have 2/8.

246
00:12:17,870 --> 00:12:20,880
For 2 we have 2/8.

247
00:12:20,880 --> 00:12:24,530
And for 3 we have 1/8.

248
00:12:24,530 --> 00:12:30,090
So this is our fully labeled
sketch of Pr of r.

249
00:12:30,090 --> 00:12:34,210
And forgive me for erasing so
quickly, but you guys can

250
00:12:34,210 --> 00:12:38,280
pause the video, presumably,
if you need more time.

251
00:12:38,280 --> 00:12:40,160
Let's move on to part d.

252
00:12:40,160 --> 00:12:43,590
So in part d we're given an
event named a, which is the

253
00:12:43,590 --> 00:12:46,890
event that x squared is greater
than or equal to y.

254
00:12:46,890 --> 00:12:50,240
And then we're asked to find the
expectation of xy in the

255
00:12:50,240 --> 00:12:51,860
unconditional universe.

256
00:12:51,860 --> 00:12:56,240
And then the expectation of x
times y conditioned on a.

257
00:12:56,240 --> 00:12:58,320
So let's not worry about the
conditioning for now.

258
00:12:58,320 --> 00:13:01,090
Let's just focus on
the unconditional

259
00:13:01,090 --> 00:13:03,040
expectation of x times y.

260
00:13:03,040 --> 00:13:04,630
So I'm just going to
erase all these r's

261
00:13:04,630 --> 00:13:07,130
so I don't get confused.

262
00:13:07,130 --> 00:13:12,910
But we're going to follow a very
similar strategy, which

263
00:13:12,910 --> 00:13:17,770
is at each point I'm going to
label what the value of w is.

264
00:13:17,770 --> 00:13:22,840
And we'll find the expectation
of w that way.

265
00:13:22,840 --> 00:13:27,130
So let's see, here,
we have 4 times 0.

266
00:13:27,130 --> 00:13:29,050
So w is equal to 0.

267
00:13:29,050 --> 00:13:30,440
Here we have 4 times 1.

268
00:13:30,440 --> 00:13:32,340
w is equal to 4.

269
00:13:32,340 --> 00:13:34,530
4 times 2, w is equal to 8.

270
00:13:34,530 --> 00:13:37,882
4 times 3, w is equal to 12.

271
00:13:37,882 --> 00:13:39,730
w is equal to 2.

272
00:13:39,730 --> 00:13:41,980
w is equal to 4.

273
00:13:41,980 --> 00:13:43,610
w is equal to 0.

274
00:13:43,610 --> 00:13:45,430
w is equal to 0.

275
00:13:45,430 --> 00:13:47,610
OK, so that was just algebra.

276
00:13:47,610 --> 00:13:51,050
And now, I claim again, we can
just write down what the

277
00:13:51,050 --> 00:13:54,450
expectation of x times y is.

278
00:13:54,450 --> 00:13:57,550
And I'm sorry, I didn't
announce my notation.

279
00:13:57,550 --> 00:14:00,560
I should mention that now.

280
00:14:00,560 --> 00:14:04,640
I was defining w to be the
random variable x times y.

281
00:14:04,640 --> 00:14:06,910
And that's why I labeled
the product of x

282
00:14:06,910 --> 00:14:10,760
times y as w over here.

283
00:14:10,760 --> 00:14:15,330
My apologies about not defining
that random variable.

284
00:14:15,330 --> 00:14:21,210
So the expectation of w, well,
w takes on a value of 0.

285
00:14:21,210 --> 00:14:23,680
When this happens, this happens
or that happens.

286
00:14:23,680 --> 00:14:28,920
And we know that those
three points occur

287
00:14:28,920 --> 00:14:31,260
with probability 3/8.

288
00:14:31,260 --> 00:14:35,100
So we have 0 times 3/8.

289
00:14:35,100 --> 00:14:38,080
I'm just using the normal
formula for expectation.

290
00:14:38,080 --> 00:14:40,815
w takes on a value of 2
with probability 1/8.

291
00:14:40,815 --> 00:14:43,590
Because this is the lead
point in which it

292
00:14:43,590 --> 00:14:46,220
happens, 2 times 1/8.

293
00:14:46,220 --> 00:14:50,770
Plus it can take on the value
of 4 with probability

294
00:14:50,770 --> 00:14:54,260
2/8, 4 times 2/8.

295
00:14:54,260 --> 00:14:58,150
And 8, with 1/8 probability.

296
00:14:58,150 --> 00:15:01,740
And similarly, 12 with
1/8 probability.

297
00:15:01,740 --> 00:15:03,150
So this is just algebra.

298
00:15:03,150 --> 00:15:06,674
The numerator sums up to 30.

299
00:15:06,674 --> 00:15:08,445
Yes, that's correct.

300
00:15:08,445 --> 00:15:13,920
So we have 30/8, which
is equal to 15/4.

301
00:15:13,920 --> 00:15:18,390
So this is our first
answer for part d.

302
00:15:18,390 --> 00:15:20,610
And now we have to do this
slightly trickier one, which

303
00:15:20,610 --> 00:15:24,290
is the conditional expectation
of x times y, or w

304
00:15:24,290 --> 00:15:26,110
conditioned on a.

305
00:15:26,110 --> 00:15:31,360
So similar to what I did in part
c, I'm going to draw the

306
00:15:31,360 --> 00:15:34,510
line y equals x squared.

307
00:15:34,510 --> 00:15:37,870
So y equals x squared
is 0 here, 1 here.

308
00:15:37,870 --> 00:15:41,070
And at 2, it should take
on a value of 4.

309
00:15:41,070 --> 00:15:45,760
So the curve should look
something like this.

310
00:15:45,760 --> 00:15:48,470
This is the line y is
equal to x squared.

311
00:15:48,470 --> 00:15:52,410
So we know all the points below
this line satisfy y less

312
00:15:52,410 --> 00:15:54,230
than or equal to x squared.

313
00:15:54,230 --> 00:15:57,130
And all the points above this
line have y greater than or

314
00:15:57,130 --> 00:15:59,170
equal to x squared.

315
00:15:59,170 --> 00:16:02,870
And a is y less than or
equal to x squared.

316
00:16:02,870 --> 00:16:07,090
So we are in the conditional
universe where only points

317
00:16:07,090 --> 00:16:09,390
below this line can happen.

318
00:16:09,390 --> 00:16:11,240
So that one, that one, that
one, that one, that

319
00:16:11,240 --> 00:16:12,190
one and that one.

320
00:16:12,190 --> 00:16:13,790
So there are six of them.

321
00:16:13,790 --> 00:16:15,930
And again, in the unconditional
world, all of

322
00:16:15,930 --> 00:16:17,190
the points were equally
likely.

323
00:16:17,190 --> 00:16:20,040
So in the conditional world
these six points are still

324
00:16:20,040 --> 00:16:21,040
equally likely.

325
00:16:21,040 --> 00:16:24,630
So they each happen with
probability 1/6.

326
00:16:24,630 --> 00:16:33,310
So in this case, the expectation
of w is simply 2

327
00:16:33,310 --> 00:16:38,050
times 1/6 plus 0 times 1/6.

328
00:16:38,050 --> 00:16:38,930
But that's 0.

329
00:16:38,930 --> 00:16:41,260
So I'm not going to write it.

330
00:16:41,260 --> 00:16:52,250
4 times 2/6 plus 4 times 2/6
plus 8 times 1/6, plus 12

331
00:16:52,250 --> 00:16:54,820
times 1 over 6.

332
00:16:54,820 --> 00:16:56,650
And again, the numerator
summed to 30.

333
00:16:56,650 --> 00:16:58,780
But this time our denominator
is 6.

334
00:16:58,780 --> 00:17:01,210
So this is simply 5.

335
00:17:01,210 --> 00:17:04,230
So we have, actually, finished
the problem.

336
00:17:04,230 --> 00:17:07,940
Because we've computed this
value and this value.

337
00:17:07,940 --> 00:17:12,369
And so the important takeaways
of this problem are,

338
00:17:12,369 --> 00:17:16,089
essentially, honestly, just to
get you comfortable with

339
00:17:16,089 --> 00:17:19,410
computing things involving
joint PMFs.

340
00:17:19,410 --> 00:17:23,109
We talked a lot about finding
expectations quickly by

341
00:17:23,109 --> 00:17:25,180
thinking about center
of mass and the

342
00:17:25,180 --> 00:17:27,030
geometry of the problem.

343
00:17:27,030 --> 00:17:31,461
We've got practice computing
conditional variances.

344
00:17:31,461 --> 00:17:34,330
And we did some derived
distributions.

345
00:17:34,330 --> 00:17:36,170
And we'll do a lot more
of those later.

346
00:17:36,170 --> 00:17:37,420