abdullah's picture
Add files using upload-large-folder tool
8de7e09 verified
raw
history blame
43.5 kB
1
00:00:08,370 --> 00:00:13,950
Today, inshallah, we'll start chapter six. Chapter
2
00:00:13,950 --> 00:00:20,350
six talks about the normal distribution. In this
3
00:00:20,350 --> 00:00:24,810
chapter, there are mainly two objectives. The
4
00:00:24,810 --> 00:00:30,470
first objective is to compute probabilities from
5
00:00:30,470 --> 00:00:34,530
normal distribution. And mainly we'll focus on
6
00:00:34,530 --> 00:00:37,270
objective number one. So we are going to use
7
00:00:37,270 --> 00:00:40,290
normal distribution in this chapter. And we'll
8
00:00:40,290 --> 00:00:43,830
know how can we compute probabilities if the data
9
00:00:43,830 --> 00:00:46,810
set is normally distributed. You know many times
10
00:00:46,810 --> 00:00:50,690
you talked about extreme points or outliers. So
11
00:00:50,690 --> 00:00:54,490
that means if the data has outliers, that is the
12
00:00:54,490 --> 00:00:57,290
distribution is not normally distributed. Now in
13
00:00:57,290 --> 00:01:01,090
this case, If the distribution is normal, how can
14
00:01:01,090 --> 00:01:04,350
we compute probabilities underneath the normal
15
00:01:04,350 --> 00:01:10,030
curve? The second objective is to use the normal
16
00:01:10,030 --> 00:01:13,210
probability plot to determine whether a set of
17
00:01:13,210 --> 00:01:18,150
data is approximately normally distributed. I mean
18
00:01:18,150 --> 00:01:25,550
beside box plots we discussed before. Beside this
19
00:01:25,550 --> 00:01:30,190
score, how can we tell if the data point or
20
00:01:30,190 --> 00:01:35,350
actually the entire distribution is approximately
21
00:01:35,350 --> 00:01:39,410
normally distributed or not. Before we learn if
22
00:01:39,410 --> 00:01:44,110
the point is outlier by using backsplot and this
23
00:01:44,110 --> 00:01:46,750
score. In this chapter we'll know how can we
24
00:01:46,750 --> 00:01:51,630
determine if the entire distribution is
25
00:01:51,630 --> 00:01:54,770
approximately normal distributed. So there are two
26
00:01:54,770 --> 00:01:56,710
objectives. One is to compute probabilities
27
00:01:56,710 --> 00:01:59,370
underneath the normal curve. The other, how can we
28
00:01:59,370 --> 00:02:05,310
tell if the data set is out or not? If you
29
00:02:05,310 --> 00:02:09,330
remember, first class, we mentioned something
30
00:02:09,330 --> 00:02:13,130
about data types. And we said data has mainly two
31
00:02:13,130 --> 00:02:17,930
types. Numerical data, I mean quantitative data.
32
00:02:18,690 --> 00:02:22,630
and categorical data, qualitative. For numerical
33
00:02:22,630 --> 00:02:26,190
data also it has two types, continuous and
34
00:02:26,190 --> 00:02:30,430
discrete. And discrete takes only integers such as
35
00:02:30,430 --> 00:02:35,310
number of students who take this class or number
36
00:02:35,310 --> 00:02:40,190
of accidents and so on. But if you are talking
37
00:02:40,190 --> 00:02:45,320
about Age, weight, scores, temperature, and so on.
38
00:02:45,560 --> 00:02:49,260
It's continuous distribution. For this type of
39
00:02:49,260 --> 00:02:53,320
variable, I mean for continuous distribution, how
40
00:02:53,320 --> 00:02:56,300
can we compute the probabilities underneath the
41
00:02:56,300 --> 00:02:59,640
normal? So normal distribution maybe is the most
42
00:02:59,640 --> 00:03:02,380
common distribution in statistics, and it's type
43
00:03:02,380 --> 00:03:07,820
of continuous distribution. So first, let's define
44
00:03:07,820 --> 00:03:12,010
continuous random variable. maybe because for
45
00:03:12,010 --> 00:03:15,230
multiple choice problem you should know the
46
00:03:15,230 --> 00:03:19,110
definition of continuous random variable is a
47
00:03:19,110 --> 00:03:22,070
variable that can assume any value on a continuous
48
00:03:23,380 --> 00:03:27,020
it can assume any uncountable number of values. So
49
00:03:27,020 --> 00:03:31,080
it could be any number in an interval. For
50
00:03:31,080 --> 00:03:35,720
example, suppose your ages range between 18 years
51
00:03:35,720 --> 00:03:39,580
and 20 years. So maybe someone of you, their age
52
00:03:39,580 --> 00:03:44,000
is about 18 years, three months. Or maybe your
53
00:03:44,000 --> 00:03:47,580
weight is 70 kilogram point five, and so on. So
54
00:03:47,580 --> 00:03:49,780
it's continuous on the variable. Other examples
55
00:03:49,780 --> 00:03:53,140
for continuous, thickness of an item. For example,
56
00:03:53,740 --> 00:03:54,440
the thickness.
57
00:03:58,260 --> 00:04:02,490
This one is called thickness. Now, the thickness
58
00:04:02,490 --> 00:04:05,930
may be 2 centimeters or 3 centimeters and so on,
59
00:04:06,210 --> 00:04:09,730
but it might be 2.5 centimeters. For example, for
60
00:04:09,730 --> 00:04:13,030
this remote, the thickness is 2.5 centimeters or 2
61
00:04:13,030 --> 00:04:16,510
.6, not exactly 2 or 3. So it could be any value.
62
00:04:16,650 --> 00:04:19,450
Range is, for example, between 2 centimeters and 3
63
00:04:19,450 --> 00:04:23,010
centimeters. So from 2 to 3 is a big range because
64
00:04:23,010 --> 00:04:25,670
it can take anywhere from 2.1 to 2.15 and so on.
65
00:04:26,130 --> 00:04:28,810
So thickness is an example of continuous random
66
00:04:28,810 --> 00:04:31,190
variable. Another example, time required to
67
00:04:31,190 --> 00:04:36,010
complete a task. Now suppose you want to do an
68
00:04:36,010 --> 00:04:39,710
exercise. Now the time required to finish or to
69
00:04:39,710 --> 00:04:45,150
complete this task may be any value between 2
70
00:04:45,150 --> 00:04:48,730
minutes up to 3 minutes. So maybe 2 minutes 30
71
00:04:48,730 --> 00:04:52,150
seconds, 2 minutes 40 seconds and so on. So it's
72
00:04:52,150 --> 00:04:55,550
continuous random variable. Temperature of a
73
00:04:55,550 --> 00:05:00,140
solution. height, weight, ages, and so on. These
74
00:05:00,140 --> 00:05:03,720
are examples of continuous random variable. So
75
00:05:03,720 --> 00:05:08,040
these variables can potentially take on any value
76
00:05:08,040 --> 00:05:11,340
depending only on the ability to precisely and
77
00:05:11,340 --> 00:05:14,020
accurately measure. So that's the definition of
78
00:05:14,020 --> 00:05:17,320
continuous random variable. Now, if you look at
79
00:05:17,320 --> 00:05:21,810
the normal distribution, It looks like bell
80
00:05:21,810 --> 00:05:25,990
-shaped, as we discussed before. So it's bell
81
00:05:25,990 --> 00:05:31,270
-shaped, symmetrical. Symmetrical means the area
82
00:05:31,270 --> 00:05:34,390
to the right of the mean equals the area to the
83
00:05:34,390 --> 00:05:37,950
left of the mean. I mean 50% of the area above and
84
00:05:37,950 --> 00:05:41,770
50% below. So that's the meaning of symmetrical.
85
00:05:42,490 --> 00:05:46,370
The other feature of normal distribution, the
86
00:05:46,370 --> 00:05:49,510
measures of center tendency are equal or
87
00:05:49,510 --> 00:05:53,170
approximately equal. Mean, median, and mode are
88
00:05:53,170 --> 00:05:55,530
roughly equal. In reality, they are not equal,
89
00:05:55,650 --> 00:05:58,210
exactly equal, but you can say they are
90
00:05:58,210 --> 00:06:01,850
approximately equal. Now, there are two parameters
91
00:06:01,850 --> 00:06:05,750
describing the normal distribution. One is called
92
00:06:05,750 --> 00:06:10,820
the location parameter. location, or central
93
00:06:10,820 --> 00:06:13,800
tendency, as we discussed before, location is
94
00:06:13,800 --> 00:06:17,160
determined by the mean mu. So the first parameter
95
00:06:17,160 --> 00:06:20,340
for the normal distribution is the mean mu. The
96
00:06:20,340 --> 00:06:24,240
other parameter measures the spread of the data,
97
00:06:24,280 --> 00:06:27,680
or the variability of the data, and the spread is
98
00:06:27,680 --> 00:06:31,860
sigma, or the variation. So we have two
99
00:06:31,860 --> 00:06:36,770
parameters, mu and sigma. The random variable in
100
00:06:36,770 --> 00:06:39,930
this case can take any value from minus infinity
101
00:06:39,930 --> 00:06:44,270
up to infinity. So random variable in this case
102
00:06:44,270 --> 00:06:50,310
continuous ranges from minus infinity all the way
103
00:06:50,310 --> 00:06:55,100
up to infinity. I mean from this point here up to
104
00:06:55,100 --> 00:06:58,380
infinity. So the values range from minus infinity
105
00:06:58,380 --> 00:07:02,080
up to infinity. And if you look here, the mean is
106
00:07:02,080 --> 00:07:05,600
located nearly in the middle. And mean and median
107
00:07:05,600 --> 00:07:10,820
are all approximately equal. That's the features
108
00:07:10,820 --> 00:07:14,740
or the characteristics of the normal distribution.
109
00:07:16,460 --> 00:07:20,360
Now, how can we compute the probabilities under
110
00:07:20,360 --> 00:07:25,840
the normal killer? The formula that is used to
111
00:07:25,840 --> 00:07:29,220
compute the probabilities is given by this one. It
112
00:07:29,220 --> 00:07:33,560
looks complicated formula because we have to use
113
00:07:33,560 --> 00:07:36,040
calculus in order to determine the area underneath
114
00:07:36,040 --> 00:07:40,120
the cube. So we are looking for something else. So
115
00:07:40,120 --> 00:07:45,300
this formula is it seems to be complicated. It's
116
00:07:45,300 --> 00:07:49,600
not hard but it's complicated one, but we can use
117
00:07:49,600 --> 00:07:52,380
it. If we know calculus very well, we can use
118
00:07:52,380 --> 00:07:55,240
integration to create the probabilities underneath
119
00:07:55,240 --> 00:07:58,900
the curve. But for our course, we are going to
120
00:07:58,900 --> 00:08:04,460
skip this formula because this
121
00:08:04,460 --> 00:08:09,340
formula depends actually on mu and sigma. A mu can
122
00:08:09,340 --> 00:08:13,110
take any value. Sigma also can take any value.
123
00:08:13,930 --> 00:08:17,310
That means we have different normal distributions.
124
00:08:18,470 --> 00:08:23,830
Because the distribution actually depends on these
125
00:08:23,830 --> 00:08:27,610
two parameters. So by varying the parameters mu
126
00:08:27,610 --> 00:08:29,790
and sigma, we obtain different normal
127
00:08:29,790 --> 00:08:32,710
distributions. Since we have different mu and
128
00:08:32,710 --> 00:08:36,310
sigma, it means we should have different normal
129
00:08:36,310 --> 00:08:38,770
distributions. For this reason, it's very
130
00:08:38,770 --> 00:08:43,430
complicated to have tables or probability tables
131
00:08:43,430 --> 00:08:46,010
in order to determine these probabilities because
132
00:08:46,010 --> 00:08:50,130
there are infinite values of mu and sigma maybe
133
00:08:50,130 --> 00:08:57,750
your edges the mean is 19. Sigma is, for example,
134
00:08:57,910 --> 00:09:01,990
5. For weights, maybe the mean is 70 kilograms,
135
00:09:02,250 --> 00:09:04,990
the average is 10. For scores, maybe the average
136
00:09:04,990 --> 00:09:08,710
is 65, the mean is 20, sigma is 20, and so on. So
137
00:09:08,710 --> 00:09:11,090
we have different values of mu and sigma. For this
138
00:09:11,090 --> 00:09:13,650
reason, we have different normal distributions.
139
00:09:18,490 --> 00:09:25,740
Because changing mu shifts the distribution either
140
00:09:25,740 --> 00:09:29,640
left or to the right. So maybe the mean is shifted
141
00:09:29,640 --> 00:09:32,440
to the right side, or the mean maybe shifted to
142
00:09:32,440 --> 00:09:37,140
the left side. Also, changing sigma, sigma is the
143
00:09:37,140 --> 00:09:40,660
distance between the mu and the curve. The curve
144
00:09:40,660 --> 00:09:45,220
is the points, or the data values. Now this sigma
145
00:09:45,220 --> 00:09:48,380
can be increases or decreases. So if sigma
146
00:09:48,380 --> 00:09:52,860
increases, it means the spread also increases. Or
147
00:09:52,860 --> 00:09:55,780
if sigma decreases, also the spread will decrease.
148
00:09:56,200 --> 00:09:59,660
So the distribution or the normal distribution
149
00:09:59,660 --> 00:10:02,820
depends actually on these two values. For this
150
00:10:02,820 --> 00:10:05,120
reason, since we have too many values or infinite
151
00:10:05,120 --> 00:10:07,600
values of mu and sigma, then in this case we have
152
00:10:07,600 --> 00:10:14,500
different normal distributions. There is another
153
00:10:14,500 --> 00:10:16,940
distribution. It's called standardized normal.
154
00:10:20,330 --> 00:10:26,070
Now, we have normal distribution X, and how can we
155
00:10:26,070 --> 00:10:31,930
transform from normal distribution to standardized
156
00:10:31,930 --> 00:10:35,310
normal distribution? The reason is that the mean
157
00:10:35,310 --> 00:10:40,310
of Z, I mean, Z is used for standardized normal.
158
00:10:40,850 --> 00:10:44,490
The mean of Z is always zero, and sigma is one.
159
00:10:45,770 --> 00:10:48,150
Now it's a big difference. The first one has
160
00:10:48,150 --> 00:10:53,160
infinite values of Mu and Sigma. Now, for the
161
00:10:53,160 --> 00:10:56,200
standardized normal distribution, the mean is
162
00:10:56,200 --> 00:11:01,540
fixed value. The mean is zero, Sigma is one. So,
163
00:11:01,620 --> 00:11:04,340
the question is, how can we actually transform
164
00:11:04,340 --> 00:11:09,720
from X, which has normal distribution, to Z, which
165
00:11:09,720 --> 00:11:13,160
has standardized normal with mean zero and Sigma
166
00:11:13,160 --> 00:11:23,330
of one. Let's see. How can we translate x which
167
00:11:23,330 --> 00:11:27,510
has normal distribution to z that has standardized
168
00:11:27,510 --> 00:11:32,190
normal distribution? The idea is you have just to
169
00:11:32,190 --> 00:11:39,170
subtract mu of x, x minus mu, then divide this
170
00:11:39,170 --> 00:11:43,150
result by sigma. So we just subtract the mean of
171
00:11:43,150 --> 00:11:49,660
x. and dividing by its standard deviation now so
172
00:11:49,660 --> 00:11:52,360
if we have x which has normal distribution with
173
00:11:52,360 --> 00:11:55,940
mean mu and standard deviation sigma to transform
174
00:11:55,940 --> 00:12:00,960
or to convert to z score use this formula x minus
175
00:12:00,960 --> 00:12:05,220
the mean then divide by its standard deviation now
176
00:12:05,220 --> 00:12:09,090
all of the time we are going to use z for
177
00:12:09,090 --> 00:12:12,230
standardized normal distribution and always z has
178
00:12:12,230 --> 00:12:15,370
mean zero and all and sigma or standard deviation.
179
00:12:16,250 --> 00:12:20,170
So the z distribution always has mean of zero and
180
00:12:20,170 --> 00:12:25,490
sigma of one. So that's the story of standardizing
181
00:12:25,490 --> 00:12:33,070
the normal value. Now the Formula for this score
182
00:12:33,070 --> 00:12:37,570
becomes better than the first one, but still we
183
00:12:37,570 --> 00:12:40,570
have to use calculus in order to determine the
184
00:12:40,570 --> 00:12:45,710
probabilities under the standardized normal k. But
185
00:12:45,710 --> 00:12:49,470
this distribution has mean of zero and sigma of
186
00:12:49,470 --> 00:12:56,910
one. So we have a table on page 570. Look at page
187
00:12:56,910 --> 00:13:00,910
570. We have table or actually there are two
188
00:13:00,910 --> 00:13:05,010
tables. One for negative value of Z and the other
189
00:13:05,010 --> 00:13:08,830
for positive value of Z. So we have two tables for
190
00:13:08,830 --> 00:13:14,730
positive and negative values of Z on page 570 and
191
00:13:14,730 --> 00:13:15,470
571.
192
00:13:17,870 --> 00:13:22,770
Now the table on page 570 looks like this one. The
193
00:13:22,770 --> 00:13:26,610
table you have starts from minus 6, then minus 5,
194
00:13:26,750 --> 00:13:32,510
minus 4.5, and so on. Here we start from minus 3.4
195
00:13:32,510 -->
223
00:16:01,880 --> 00:16:06,470
mean of x is 100. and the standard deviation of
224
00:16:06,470 --> 00:16:11,110
50. So sigma is 50. Now let's see how can we
225
00:16:11,110 --> 00:16:17,750
compute the z-score for x equals 200. Again the
226
00:16:17,750 --> 00:16:22,790
formula is just x minus mu divided by sigma x 200
227
00:16:22,790 --> 00:16:28,330
minus 100 divided by 50 that will give 2. Now the
228
00:16:28,330 --> 00:16:33,910
sign of this value is positive. That means x is
229
00:16:33,910 --> 00:16:37,950
greater than the mean, because x is 200. Now,
230
00:16:37,990 --> 00:16:42,270
what's the meaning of 2? What does this value tell
231
00:16:42,270 --> 00:16:42,410
you?
232
00:16:48,230 --> 00:16:55,430
Yeah, exactly. x equals 200 is two standard
233
00:16:55,430 --> 00:16:58,690
deviations above the mean. Because if you look at
234
00:16:58,690 --> 00:17:05,210
200, the x value, the mean is 100, sigma is 50.
235
00:17:05,730 --> 00:17:09,690
Now the difference between the score, which is
236
00:17:09,690 --> 00:17:16,810
200, and the mu, which is 100, is equal to
237
00:17:16,810 --> 00:17:18,690
standard deviations, because the difference is
238
00:17:18,690 --> 00:17:24,230
100. 2 times 50 is 100. So this says that x equals
239
00:17:24,230 --> 00:17:29,070
200 is 2 standard deviations above the mean. If z
240
00:17:29,070 --> 00:17:34,330
is negative, you can say that x is two standard
241
00:17:34,330 --> 00:17:38,710
deviations below the mean. Make sense? So that's how
242
00:17:38,710 --> 00:17:42,670
we can compute the z score. Now, when we
243
00:17:42,670 --> 00:17:45,970
transform from normal distribution to
244
00:17:45,970 --> 00:17:49,490
standardized, still we will have the same shape. I
245
00:17:49,490 --> 00:17:51,350
mean the distribution is still normally
246
00:17:51,350 --> 00:17:55,800
distributed. So note, the shape of the
247
00:17:55,800 --> 00:17:58,840
distribution is the same, only the scale has
248
00:17:58,840 --> 00:18:04,500
changed. So we can express the problem in original
249
00:18:04,500 --> 00:18:10,640
units, X, or in a standardized unit, Z. So when we
250
00:18:10,640 --> 00:18:16,620
have X, just use this equation to transform to
251
00:18:16,620 --> 00:18:17,160
this form.
252
00:18:21,360 --> 00:18:23,200
Now, for example, suppose we have a normal
253
00:18:23,200 --> 00:18:26,040
distribution and we are interested in the area
254
00:18:26,040 --> 00:18:32,660
between A and B. Now, the area between A and B, it
255
00:18:32,660 --> 00:18:34,700
means the probability between them. So
256
00:18:34,700 --> 00:18:39,140
statistically speaking, area means probability. So
257
00:18:39,140 --> 00:18:42,700
probability between A and B, I mean probability of
258
00:18:42,700 --> 00:18:45,380
X greater than or equal A and less than or equal B
259
00:18:45,380 --> 00:18:49,420
is the same as X greater than A or less than B.
260
00:18:50,450 --> 00:18:57,210
That means the probability of X equals A this
261
00:18:57,210 --> 00:19:02,510
probability is zero or probability of X equals B
262
00:19:02,510 --> 00:19:06,930
is also zero. So in continuous distribution the
263
00:19:06,930 --> 00:19:10,630
equal sign does not matter, I mean if we have equal
264
00:19:10,630 --> 00:19:15,130
sign or we don't have these probabilities are the
265
00:19:15,130 --> 00:19:19,390
same, so I mean, for example if we are interested
266
00:19:20,310 --> 00:19:23,450
for probability of X smaller than or equal to E.
267
00:19:24,850 --> 00:19:30,370
This probability is the same as X smaller than E.
268
00:19:31,330 --> 00:19:33,730
Or on the other hand, if you are interested in the
269
00:19:33,730 --> 00:19:39,010
area above B greater than or equal to B, it's the
270
00:19:39,010 --> 00:19:44,770
same as X smaller than E. So don't worry about the
271
00:19:44,770 --> 00:19:48,660
equal sign. For continuous distribution, exactly.
272
00:19:49,120 --> 00:19:53,820
But for discrete, it does matter. Now, since we
273
00:19:53,820 --> 00:19:58,200
are talking about normal distribution, and as we
274
00:19:58,200 --> 00:20:01,320
mentioned, normal distribution is symmetric around
275
00:20:01,320 --> 00:20:05,900
the mean, that means the area to the right equals
276
00:20:05,900 --> 00:20:09,340
the area to the left. Now the entire area
277
00:20:09,340 --> 00:20:12,940
underneath the normal curve equals one. I mean
278
00:20:12,940 --> 00:20:16,500
probability of X ranges from minus infinity up to
279
00:20:16,500 --> 00:20:21,500
infinity equals one. So probability of X greater
280
00:20:21,500 --> 00:20:26,920
than minus infinity up to infinity is one. The
281
00:20:26,920 --> 00:20:31,480
total area is one. So the area from minus infinity
282
00:20:31,480 --> 00:20:38,080
up to the mean mu is one-half. The same as the
283
00:20:38,080 --> 00:20:42,600
area from mu up to infinity is also one-half. That
284
00:20:42,600 --> 00:20:44,760
means the probability of X greater than minus
285
00:20:44,760 --> 00:20:48,300
infinity up to mu equals the probability from mu
286
00:20:48,300 --> 00:20:52,120
up to infinity because of symmetry. I mean you
287
00:20:52,120 --> 00:20:56,160
cannot say that for any distribution. Just for
288
00:20:56,160 --> 00:20:59,000
symmetric distribution, the area below the mean
289
00:20:59,000 --> 00:21:03,780
equals one-half, which is the same as the area to
290
00:21:03,780 --> 00:21:07,110
the right of the mean. So the entire probability
291
00:21:07,110 --> 00:21:11,330
is one. And also you have to keep in mind that the
292
00:21:11,330 --> 00:21:17,570
probability always ranges between zero and one. So
293
00:21:17,570 --> 00:21:20,030
that means the probability couldn't be negative.
294
00:21:22,870 --> 00:21:27,730
It should be positive. It shouldn't be greater
295
00:21:27,730 --> 00:21:31,710
than one. So it's between zero and one. So always
296
00:21:31,710 --> 00:21:39,020
the probability lies between zero and one. The
297
00:21:39,020 --> 00:21:44,500
tables we have on page 570 and 571 give the area
298
00:21:44,500 --> 00:21:46,040
to the left side.
299
00:21:49,420 --> 00:21:54,660
For negative or positive z's. Now for example,
300
00:21:54,940 --> 00:22:03,060
suppose we are looking for probability of z less
301
00:22:03,060 --> 00:22:08,750
than 2. How can we find this probability by using
302
00:22:08,750 --> 00:22:12,210
the normal curve? Let's go back to this normal
303
00:22:12,210 --> 00:22:16,410
distribution. In the second page, we have positive
304
00:22:16,410 --> 00:22:17,070
z-scores.
305
00:22:23,850 --> 00:22:33,390
So we ask about the probability of z less than. So
306
00:22:33,390 --> 00:22:40,690
the second page, gives positive values of z. And
307
00:22:40,690 --> 00:22:44,590
the table gives the area below. And he asked about
308
00:22:44,590 --> 00:22:49,550
here, B of z is smaller than 2. Now 2, if you
309
00:22:49,550 --> 00:22:54,910
hear, up all the way down here, 2, 0, 0. So the
310
00:22:54,910 --> 00:23:00,530
answer is 9772. So this value, so the probability
311
00:23:00,530 --> 00:23:02,130
is 9772.
312
00:23:03,990 --> 00:23:05,390
Because it's 2.
313
00:23:09,510 --> 00:23:14,650
It's 2, 0, 0. But if you ask about what's the
314
00:23:14,650 --> 00:23:20,590
probability of Z less than 2.05? So this is 2.
315
00:23:23,810 --> 00:23:30,370
Now under 5, 9, 7, 9, 8. So the answer is 9, 7.
316
00:23:34,360 --> 00:23:38,900
Because this is two, and we need five decimal
317
00:23:38,900 --> 00:23:44,820
places. So all the way up to 9798. So this value
318
00:23:44,820 --> 00:23:54,380
is 2.05. Now it's about, it's more than 1.5,
319
00:23:55,600 --> 00:23:56,880
exactly 1.5.
320
00:24:02,140 --> 00:24:04,880
1.5. This is 1.5.
321
00:24:08,800 --> 00:24:09,720
9332.
322
00:24:12,440 --> 00:24:16,300
1.5. Exactly 1.5. So 9332.
323
00:24:18,780 --> 00:24:27,990
What's about probability less than 1.35? 1.3 all
324
00:24:27,990 --> 00:24:35,250
the way to 9.115. 9.115. 9.115. 9.115. 9.115. 9
325
00:24:35,250 --> 00:24:35,650
.115. 9.115.
326
00:24:41,170 --> 00:24:42,430
9.115. 9.115. 9.115. 9.115. 9.115. 9.115. 9.115. 9
327
00:24:42,430 --> 00:24:42,450
.115. 9.115. 9.115. 9.115. 9.115. 9.115. 9.115. 9
328
00:24:42,450 --> 00:24:44,050
.115. 9.115. 9.115. 9.115. 9.115. 9.115. 9.115. 9
329
00:24:44,050 --> 00:24:50,530
.115. 9.115. 9.115. 9.115. 9.115. 9.115. 9.115. 9
330
00:24:50,530 --> 00:24:54,980
.115. 9. But here we are looking for the area to
331
00:24:54,980 --> 00:25:01,280
the right. One minus one. Now this area equals
332
00:25:01,280 --> 00:25:05,660
one minus because
333
00:25:05,660 --> 00:25:11,420
since suppose
334
00:25:11,420 --> 00:25:18,760
this is the 1.35 and we are interested in the area
335
00:25:18,760 --> 00:25:24,030
to the right or above 1.35. The table gives the
336
00:25:24,030 --> 00:25:28,230
area below. So the area above equals the total
337
00:25:28,230 --> 00:25:31,970
area underneath the curve is 1. So 1 minus this
338
00:25:31,970 --> 00:25:39,050
value, so equals 0.0885,
339
00:25:39,350 --> 00:25:42,250
and so on. So this is the way how can we compute
340
00:25:42,250 --> 00:25:47,850
the probabilities underneath the normal curve. If
341
00:25:47,850 --> 00:25:51,090
it's probability of z is smaller than then just
342
00:25:51,090 --> 00:25:55,910
use the table directly otherwise if we are talking
343
00:25:55,910 --> 00:26:00,390
about z greater than subtract from one to get the
344
00:26:00,390 --> 00:26:04,870
result. That's how can we compute the probability
345
00:26:04,870 --> 00:26:13,750
of z less than or equal. Now
346
00:26:13,750 --> 00:26:18,890
let's see if we have x and x that has a normal
347
00:26:18,890 --> 00:26:22,070
distribution with mean mu and standard deviation
348
00:26:22,070 --> 00:26:26,250
of sigma and let's see how can we compute the
349
00:26:26,250 --> 00:26:33,790
value of the probability. Mainly
350
00:26:33,790 --> 00:26:38,190
there are three steps to find the probability of x
351
00:26:38,190 --> 00:26:42,490
greater than a and less than b when x is
352
00:26:42,490 --> 00:26:47,000
distributed normally. First step. Draw a normal curve
353
00:26:47,000 --> 00:26:54,880
for the problem in terms of x. So draw the normal
354
00:26:54,880 --> 00:26:58,140
curve first. Second, translate x values to z
355
00:26:58,140 --> 00:27:03,040
values by using the formula we have. z equals x minus mu
356
00:27:03,040 --> 00:27:06,440
divided by sigma. Then use the standardized normal
357
00:27:06,440 --> 00:27:15,140
table on page 570 and 571. For example, let's see
358
00:27:15,140 --> 00:27:18,420
how can we find normal probabilities. Let's assume
359
00:27:18,420 --> 00:27:23,760
that X represents the time it takes to download an
360
00:27:23,760 --> 00:27:28,580
image from the internet. So suppose X, time
361
00:27:28,580 --> 00:27:33,760
required to download an image file from the
362
00:27:33,760 --> 00:27:38,460
internet. And suppose we know that the time is
363
00:27:38,460 --> 00:27:42,060
normally distributed with a mean of eight
364
00:27:42,060 --> 00:27:46,130
minutes and standard deviation of five minutes.
365
00:27:46,490 --> 00:27:47,510
So we know the mean.
366
00:27:50,610 --> 00:27:59,670
Eight. Eight. And sigma of five minutes. And they
367
00:27:59,670 --> 00:28:03,410
ask about what's the probability of X smaller than
368
00:28:03,410 --> 00:28:07,990
8.16. So first thing we have to compute,
369
00:28:08,170 --> 00:28:12,190
to draw the normal curve. The mean lies in the
370
00:28:12,190 --> 00:28:18,060
center, which is 8. He asked about probability of
371
00:28:18,060 --> 00:28:22,580
X smaller than 8.6. So we are interested in the
372
00:28:22,580 --> 00:28:27,920
area below 8.6. So it matched the table we have.
373
00:28:29,980 --> 00:28:34,900
Second step, we have to transform from normal
374
00:28:34,900 --> 00:28:37,280
distribution to standardized normal distribution
375
00:28:37,280 --> 00:28:42,120
by using this form, which is X minus mu divided by
376
00:28:42,120 --> 00:28:51,430
sigma. So x is 8.6 minus the mean, 8, divided by
377
00:28:51,430 --> 00:28:57,130
sigma, gives 0.12. So just straightforward
378
00:28:57,130 --> 00:29:02,890
calculation, 8.6 is your value of x. The mean is
379
00:29:02,890 --> 00:29:12,810
8, sigma is 5, so that gives 0.12. So now, the
380
00:29:12,810 --> 00:29:17,210
problem becomes, instead of asking x smaller than
381
00:29:17,210 --> 00:29:25,110
8.6, it's similar to z less than 0.12. Still, we
382
00:29:25,110 --> 00:29:26,310
have the same normal curve.
383
00:29:29,450 --> 00:29:32,990
8, the mean. Now, the mean of z is 0, as we
384
00:29:32,990 --> 00:29:39,230
mentioned. Instead of x, 8.6, the corresponding z
385
00:29:39,230 --> 00:29:43,000
value is 0.12. So instead of finding probability
386
00:29:43,000 --> 00:29:48,580
of X smaller than 8.6, smaller than 1.12, so they
387
00:29:48,580 --> 00:29:53,760
are equivalent. So we transform here from normal
388
00:29:53,760 --> 00:29:56,980
distribution to standardized normal distribution
389
00:29:56,980 --> 00:29:59,980
in order to compute the probability we are looking
390
00:29:59,980 --> 00:30:05,820
for. Now, this is just a portion of the table we
391
00:30:05,820 --> 00:30:06,100
have.
392
00:30:10,530 --> 00:30:18,530
So for positive z values. Now 0.1 is 0.1. Because
393
00:30:18,530 --> 00:30:25,670
here we are looking for z less than 0.1. So 0.1.
394
00:30:27,210 --> 00:30:32,950
Also, we have two. So move up to two decimal
395
00:30:32,950 --> 00:30:38,190
places, we get this value. So the answer is point.
396
00:30:42,120 --> 00:30:45,860
I think it's straightforward to compute the
397
00:30:45,860 --> 00:30:49,460
probability underneath the normal curve if X has
398
00:30:49,460 --> 00:30:53,160
normal distribution. So B of X is smaller than 8.6
399
00:30:53,160 --> 00:30:56,740
is the same as B of Z less than 0.12, which is
400
00:30:56,740 --> 00:31:02,680
around 55%. Makes sense because the area to the
401
00:31:02,680 --> 00:31:07,080
left of 0 equals 1 half. But we are looking for
402
00:31:07,080 --> 00:31:12,440
the area below 0.12. So greater than zero. So this
403
00:31:12,440 --> 00:31:16,600
area actually is greater than 0.5. So it makes
404
00:31:16,600 --> 00:31:20,440
sense that your result is greater than 0.5.
405
00:31:22,320 --> 00:31:22,960
Questions?
406
00:31:25,480 --> 00:31:30,780
Next, suppose we are interested in
445
00:34:59,320 --> 00:35:04,240
answer is going to be 0.478. So that's how can we
446
00:35:04,240 --> 00:35:07,540
compute the probabilities for lower 10 directly
447
00:35:07,540 --> 00:35:12,230
from the table. upper tail is just one minus lower
448
00:35:12,230 --> 00:35:18,990
tail and between two values just subtracts the
449
00:35:18,990 --> 00:35:21,970
larger one minus smaller one because he was
450
00:35:21,970 --> 00:35:26,310
subtracted by less than point one minus bz less
451
00:35:26,310 --> 00:35:29,430
than or equal to zero that will give the normal
452
00:35:29,430 --> 00:35:36,850
probability another example suppose we are looking
453
00:35:36,850 --> 00:35:49,350
for X between 7.4 and 8. Now, 7.4 lies below the
454
00:35:49,350 --> 00:35:55,270
mean. So here, this value, we have to compute the
455
00:35:55,270 --> 00:36:00,130
z-score for 7.4 and also the z-score for 8, which
456
00:36:00,130 --> 00:36:04,090
is zero. And that will give, again,
457
00:36:07,050 --> 00:36:13,710
7.4, if you just use this equation, minus
458
00:36:13,710 --> 00:36:17,690
the mean, divided by sigma, negative 0.6 divided
459
00:36:17,690 --> 00:36:21,150
by 5, which is negative 0.12.
460
00:36:22,730 --> 00:36:31,410
So it gives B of z between minus 0.12 and 0. And
461
00:36:31,410 --> 00:36:35,700
that again is B of z less than 0. minus P of Z
462
00:36:35,700 --> 00:36:40,140
less than negative 0.12. Is it clear? Now here we
463
00:36:40,140 --> 00:36:42,260
converted or we transformed from normal
464
00:36:42,260 --> 00:36:45,960
distribution to standardized. So instead of X
465
00:36:45,960 --> 00:36:52,100
between 7.4 and 8, we have now Z between minus 0
466
00:36:52,100 --> 00:36:57,480
.12 and 0. So this area actually is the red one,
467
00:36:57,620 --> 00:37:03,740
the red area is one-half. Total area below z is
468
00:37:03,740 --> 00:37:10,700
one-half, below zero, and minus z below minus 0
469
00:37:10,700 --> 00:37:17,820
.12. So B of z less than zero minus negative 0.12.
470
00:37:18,340 --> 00:37:21,940
That will give the area between minus 0.12 and
471
00:37:21,940 --> 00:37:28,860
zero. This is one-half. Now, B of z less than
472
00:37:28,860 --> 00:37:33,270
negative 0.12. look you go back to the normal
473
00:37:33,270 --> 00:37:37,650
curve to the normal table but for the negative
474
00:37:37,650 --> 00:37:42,310
values of z negative point one two negative point
475
00:37:42,310 --> 00:37:53,290
one two four five two two it's four five point
476
00:37:53,290 --> 00:37:56,630
five minus point four five two two will give the
477
00:37:56,630 --> 00:37:58,370
result we are looking for
478
00:38:01,570 --> 00:38:06,370
So B of Z less than 0 is 0.5. B of Z less than
479
00:38:06,370 --> 00:38:12,650
negative 0.12 equals minus 0.4522. That will give
480
00:38:12,650 --> 00:38:14,290
0 forcibility.
481
00:38:16,790 --> 00:38:23,590
Now, by symmetric, you can see that this
482
00:38:23,590 --> 00:38:28,470
probability between
483
00:38:28,470 --> 00:38:38,300
Z between minus 0.12 and 0 is the same as the
484
00:38:38,300 --> 00:38:43,340
other side from 0.12 I mean this area the red one
485
00:38:43,340 --> 00:38:46,200
is the same up to 8.6
486
00:38:55,600 --> 00:38:58,840
So the area between minus 0.12 up to 0 is the same
487
00:38:58,840 --> 00:39:04,920
as from 0 up to 0.12. Because of symmetric, since
488
00:39:04,920 --> 00:39:09,680
this area equals the same for the other part. So
489
00:39:09,680 --> 00:39:15,660
from 0 up to 0.12 is the same as minus 0.12 up to
490
00:39:15,660 --> 00:39:19,100
0. So equal, so the normal distribution is
491
00:39:19,100 --> 00:39:23,200
symmetric. So this probability is the same as B of
492
00:39:23,200 --> 00:39:27,980
Z between 0 and 0.12. Any question?
493
00:39:34,520 --> 00:39:36,620
Again, the equal sign does not matter.
494
00:39:42,120 --> 00:39:45,000
Because here we have the complement. The
495
00:39:45,000 --> 00:39:49,250
complement. If this one, I mean, complement of z
496
00:39:49,250 --> 00:39:53,350
less than, greater than 0.12, the complement is B
497
00:39:53,350 --> 00:39:56,350
of z less than or equal to minus 0.12. So we
498
00:39:56,350 --> 00:40:00,070
should have just permutation, the equality. But it
499
00:40:00,070 --> 00:40:04,830
doesn't matter. If in the problem we don't have
500
00:40:04,830 --> 00:40:07,470
equal sign in the complement, we should have equal
501
00:40:07,470 --> 00:40:11,430
sign. But it doesn't matter actually if we have
502
00:40:11,430 --> 00:40:14,510
equal sign or not. For example, if we are looking
503
00:40:14,510 --> 00:40:19,430
for B of X greater than A. Now what's the
504
00:40:19,430 --> 00:40:25,950
complement of that? 1 minus less
505
00:40:25,950 --> 00:40:32,450
than or equal to A. But if X is greater than or
506
00:40:32,450 --> 00:40:37,870
equal to A, the complement is without equal sign.
507
00:40:38,310 --> 00:40:40,970
But in continuous distribution, the equal sign
508
00:40:40,970 --> 00:40:44,990
does not matter. Any question?
509
00:40:52,190 --> 00:40:58,130
comments. Let's move to the next topic which talks
510
00:40:58,130 --> 00:41:05,510
about the empirical rule. If you remember before
511
00:41:05,510 --> 00:41:16,750
we said there is an empirical rule for 68, 95, 95,
512
00:41:17,420 --> 00:41:23,060
99.71. Now let's see the exact meaning of this
513
00:41:23,060 --> 00:41:23,320
rule.
514
00:41:37,580 --> 00:41:40,460
Now we have to apply the empirical rule not to
515
00:41:40,460 --> 00:41:43,020
Chebyshev's inequality because the distribution is
516
00:41:43,020 --> 00:41:48,670
normal. Chebyshev's is applied for skewed
517
00:41:48,670 --> 00:41:52,630
distributions. For symmetric, we have to apply the
518
00:41:52,630 --> 00:41:55,630
empirical rule. Here, we assume the distribution
519
00:41:55,630 --> 00:41:58,390
is normal. And today, we are talking about normal
520
00:41:58,390 --> 00:42:01,330
distribution. So we have to use the empirical
521
00:42:01,330 --> 00:42:02,410
rules.
522
00:42:07,910 --> 00:42:13,530
Now, the mean is the value in the middle. Suppose
523
00:42:13,530 --> 00:42:16,900
we are far away from the mean by one standard
524
00:42:16,900 --> 00:42:22,720
deviation either below or above and we are
525
00:42:22,720 --> 00:42:27,040
interested in the area between this value which is
526
00:42:27,040 --> 00:42:33,040
mu minus sigma so we are looking for mu minus
527
00:42:33,040 --> 00:42:36,360
sigma and mu plus sigma
528
00:42:53,270 --> 00:42:59,890
Last time we said there's a rule 68% of the data
529
00:42:59,890 --> 00:43:06,790
lies one standard deviation within the mean. Now
530
00:43:06,790 --> 00:43:10,550
let's see how can we compute the exact area, area
531
00:43:10,550 --> 00:43:15,250
not just say 68%. Now X has normal distribution
532
00:43:15,250 --> 00:43:18,390
with mean mu and standard deviation sigma. So
533
00:43:18,390 --> 00:43:25,280
let's compare it from normal distribution to
534
00:43:25,280 --> 00:43:29,700
standardized. So this is the first value here. Now
535
00:43:29,700 --> 00:43:34,940
the z-score, the general formula is x minus the
536
00:43:34,940 --> 00:43:40,120
mean divided by sigma. Now the first quantity is
537
00:43:40,120 --> 00:43:45,660
mu minus sigma. So instead of x here, so first z
538
00:43:45,660 --> 00:43:49,820
is, now this x should be replaced by mu minus
539
00:43:49,820 --> 00:43:55,040
sigma. So mu minus sigma. So that's my x value,
540
00:43:55,560 --> 00:44:00,240
minus the mean of that, which is mu, divided by
541
00:44:00,240 --> 00:44:07,900
sigma. Mu minus sigma minus mu mu cancels, so plus
542
00:44:07,900 --> 00:44:13,520
one. And let's see how can we compute that area. I
543
00:44:13,520 --> 00:44:16,980
mean between minus one and plus one. In this case,
544
00:44:17,040 --> 00:44:23,180
we are interested or we are looking for the area
545
00:44:23,180 --> 00:44:28,300
between minus one and plus one this area now the
546
00:44:28,300 --> 00:44:31,360
dashed area I mean the area between minus one and
547
00:44:31,360 --> 00:44:39,460
plus one equals the area below one this area minus
548
00:44:39,460 --> 00:44:44,980
the area below minus one that will give the area
549
00:44:44,980 --> 00:44:48,200
between minus one and plus one now go back to the
550
00:44:48,200 --> 00:44:52,500
normal table you have and look at the value of one
551
00:44:52,500 --> 00:45:02,620
z and one under zero what's your answer one point
552
00:45:02,620 --> 00:45:11,520
one point now without using the table can you tell
553
00:45:11,520 --> 00:45:17,360
the area below minus one one minus this one
554
00:45:17,360 --> 00:45:17,840
because
555
00:45:23,430 --> 00:45:29,870
Now the area below, this is 1. The area below 1 is
556
00:45:29,870 --> 00:45:31,310
0.3413.
557
00:45:34,430 --> 00:45:37,590
Okay, now the area below minus 1.
558
00:45:40,770 --> 00:45:42,050
This is minus 1.
559
00:45:46,810 --> 00:45:49,550
Now, the area below minus 1 is the same as above
560
00:45:49,550 --> 00:45:50,510
1.
561
00:45:54,310 --> 00:45:58,810
These are the two areas here are equal. So the
562
00:45:58,810 --> 00:46:03,110
area below minus 1, I mean b of z less than minus
563
00:46:03,110 --> 00:46:09,130
1 is the same as b of z greater than 1. And b of z
564
00:46:09,130 --> 00:46:12,650
greater than 1 is the same as 1 minus b of z
565
00:46:12,650 --> 00:46:17,310
smaller than 1. So b of z less than 1 here. You
566
00:46:17,310 --> 00:46:19,710
shouldn't need to look again to the table. Just
567
00:46:19,710 --> 00:46:26,770
subtract 1 from this value. Make sense? Here we
568
00:46:26,770 --> 00:46:30,490
compute the value of B of Z less than 1, which is
569
00:46:30,490 --> 00:46:35,430
0.8413. We are looking for B of Z less than minus
570
00:46:35,430 --> 00:46:39,770
1, which is the same as B of Z greater than 1.
571
00:46:40,750 --> 00:46:43,850
Now, greater than means our tail. It's 1 minus the
572
00:46:43,850 --> 00:46:48,700
lower tail probability. So this is 1 minus. So the
573
00:46:48,700 --> 00:46:52,240
answer again is 1 minus 0.8413.
574
00:46:54,280 --> 00:47:00,040
So 8413 minus 0.1587.
575
00:47:11,380 --> 00:47:17,030
So 8413. minus 1.1587.
576
00:47:21,630 --> 00:47:27,570
Okay, so that gives 0.6826.
577
00:47:29,090 --> 00:47:37,550
Multiply this one by 100, we get 68.1826.
578
00:47:38,750 --> 00:47:44,010
So roughly 60-80% of the observations lie between
579
00:47:44,010 --> 00:47:50,470
one standard deviation around the mean. So this is
580
00:47:50,470 --> 00:47:53,850
the way how can we compute the area below one
581
00:47:53,850 --> 00:47:57,250
standard deviation or above one standard deviation
582
00:47:57,250 --> 00:48:03,790
of the mean. Do the same for not mu minus sigma,
583
00:48:05,230 --> 00:48:11,540
mu plus minus two sigma and mu plus two sigma. The
584
00:48:11,540 --> 00:48:14,600
only difference is that this one is going to be
585
00:48:14,600 --> 00:48:17,280
minus 2 and do the same.
586
00:48:20,620 --> 00:48:23,080
That's the empirical rule we discussed in chapter
587
00:48:23,080 --> 00:48:28,980
3. So here we can find any probability, not just
588
00:48:28,980 --> 00:48:33,660
95 or 68 or 99.7. We can use the normal table to
589
00:48:33,660 --> 00:48:36,900
give or to find or to compute any probability.
590
00:48:48,270 --> 00:48:53,090
So again, for the other one, mu plus or minus two
591
00:48:53,090 --> 00:49:00,190
sigma, it covers about 95% of the axis. For mu
592
00:49:00,190 --> 00:49:03,750
plus or minus three sigma, it covers around all
593
00:49:03,750 --> 00:49:08,450
the data, 99.7. So just do it at home, you will
594
00:49:08,450 --> 00:49:14,210
see that the exact area is 95.44 instead of 95.
595
00:49:14,840 --> 00:49:18,520
And the other one is 99.73. So that's the
596
00:49:18,520 --> 00:49:23,520
empirical rule we discussed in chapter three. I'm
597
00:49:23,520 --> 00:49:32,560
going to stop at this point, which is the x value
598
00:49:32,560 --> 00:49:38,400
for the normal probability. Now, what we discussed
599
00:49:38,400 --> 00:49:43,560
so far, we computed the probability. I mean,
600
00:49:43,740 --> 00:49:49,120
what's the probability of X smaller than E? Now,
601
00:49:49,200 --> 00:49:56,240
suppose this probability is known. How can we
602
00:49:56,240 --> 00:50:01,500
compute this value? Later, we'll talk about that.
603
00:50:06,300 --> 00:50:09,820
It's backward calculations. It's inverse or
604
00:50:09,820 --> 00:50:11,420
backward calculation.
605
00:50:13,300 --> 00:50:14,460
for next time inshallah.