Professional Documents
Culture Documents
2
00:00:15,500 --> 00:00:19,680
Welcome again to our this video lecture on
basic surveying.
3
00:00:21,700 --> 00:00:28,919
Today, we are in module 2 and in lecture number
4, and the name of the module is ‘Basic
4
00:00:28,920 --> 00:00:30,420
Concepts of Surveying’.
5
00:00:32,560 --> 00:00:40,560
Well, in the entire structure of our this
video lecture, we are here at the moment in
6
00:00:40,570 --> 00:00:42,640
the module number 2.
7
00:00:42,640 --> 00:00:51,110
What we have done so far, we have discussed lesson number 1, 2 and 3, and there in
the
8
00:00:51,110 --> 00:00:57,559
three, we talked about various things, particularly about the errors in the
measurement, because
9
00:00:57,559 --> 00:01:02,959
surveying, as we said, is about measurement,
and whenever we are taking a measurement,
10
00:01:02,960 --> 00:01:06,940
we are introducing - there is some errors
coming in to the observation.
11
00:01:07,820 --> 00:01:10,100
So, we need to understand about those errors.
12
00:01:10,740 --> 00:01:15,820
What are their sources - this is what we have discussed - we saw the sources could
be natural,
13
00:01:15,830 --> 00:01:21,470
could be the human-induced - the personal
one, also the instruments which we are using
14
00:01:21,470 --> 00:01:22,710
may result in some kind of errors.
15
00:01:23,120 --> 00:01:24,900
We saw some examples also of that.
16
00:01:25,640 --> 00:01:31,340
Then in the classification of errors, the
errors could be, you know, the blunders which
17
00:01:31,350 --> 00:01:38,729
we commit - we saw it - instead of writing,
maybe, let us say, 31.5, what I am writing,
18
00:01:38,730 --> 00:01:40,200
I am writing 3.15.
19
00:01:40,200 --> 00:01:45,350
So, things like this may happen, and these
are the blunders; mistakes.
20
00:01:45,350 --> 00:01:53,119
Then, the second one: there are various physical
forces which are acting - the temperature
21
00:01:53,119 --> 00:01:59,520
may be more, the gravity is working, the wind
is blowing, and many of these factors, they
22
00:01:59,520 --> 00:02:02,960
contribute some errors, again, in the observation.
23
00:02:02,960 --> 00:02:08,150
But about these errors, we can quantify them;
we can write a mathematical model; you can
24
00:02:08,150 --> 00:02:11,420
write - because you know the physical law
behind these errors.
25
00:02:11,420 --> 00:02:14,660
So, from our observations, we can eliminate
those errors.
26
00:02:14,660 --> 00:02:22,540
These errors which can be eliminated by writing
a mathematical law are called systematic errors.
27
00:02:22,540 --> 00:02:26,560
Then, next we discussed was about the random
errors.
28
00:02:26,560 --> 00:02:34,650
From your observation side, you have eliminated
mistakes or blunders - taken away them.
29
00:02:34,650 --> 00:02:40,170
We have also taken away the systematic error
component, but still your observations are
30
00:02:40,170 --> 00:02:42,260
not true values.
31
00:02:42,260 --> 00:02:45,700
We saw that it is never possible to observe
the true value.
32
00:02:45,700 --> 00:02:46,738
So what we saw?
33
00:02:46,739 --> 00:02:52,390
At this stage, the errors which are there
in the observations are random errors.
34
00:02:52,390 --> 00:02:57,480
We saw the distribution, what kind of distribution
is there for the random errors - they are
35
00:02:57,480 --> 00:03:05,280
equally in positive and negative sides; small
errors, the small errors have got high probability,
36
00:03:05,280 --> 00:03:10,150
while the large errors on the side of the
curve - we saw it - has got less probability.
37
00:03:10,150 --> 00:03:17,599
Then we talked about principle of adjustment,
when the errors are there in the observations
38
00:03:17,599 --> 00:03:20,410
many times, you know, our network there in
the field.
39
00:03:20,410 --> 00:03:22,769
For example, we had an example of the triangle.
40
00:03:23,440 --> 00:03:29,079
So, the geometry from the field says the sum of those three angles in a triangle
which
41
00:03:29,080 --> 00:03:37,550
you have measured should be 180, but we saw it is not so, because in each and every
theta
42
00:03:37,550 --> 00:03:40,210
which we measured - the angle - we had some error.
43
00:03:41,160 --> 00:03:47,480
So, we need to adjust our observations so
that they conform to the geometry which is
44
00:03:47,480 --> 00:03:48,480
there in the field.
45
00:03:50,040 --> 00:03:55,600
Well, we did not talk about this principle
of reversal - we will talk about this principle
46
00:03:55,610 --> 00:03:58,790
of reversal later on, when we are talking
about some instruments; adjustment of the
47
00:03:58,790 --> 00:03:59,790
instruments.
48
00:04:00,500 --> 00:04:09,439
Then, we saw some more basic terms: the true
error; the percentage error - relative error;
49
00:04:09,439 --> 00:04:14,989
one very important concept we looked at was
the most probable value - a most probable
50
00:04:14,989 --> 00:04:20,340
value is a value which has maximum chances
of being nearest to the true value.
51
00:04:20,339 --> 00:04:25,940
Then, we saw how we can find it; we proved
it that the most probable value for a set
52
00:04:25,940 --> 00:04:28,250
of observations is the arithmetic mean.
53
00:04:29,520 --> 00:04:34,320
Well, towards the end of the lecture, we were
looking at the random error distribution,
54
00:04:34,330 --> 00:04:42,050
and we saw that the random errors are normally distributed; we saw the bell-type
curve.
55
00:04:42,840 --> 00:04:47,340
So, what we will do today in our today’s
lecture, we will again look into a little
56
00:04:47,340 --> 00:04:53,741
detail of random error distribution, while
mostly, we will be talking about use of normal
57
00:04:53,741 --> 00:04:59,861
distribution curve because we know our errors follow this pattern; we can assume it
very
58
00:04:59,861 --> 00:05:00,601
safely.
59
00:05:00,860 --> 00:05:07,240
So, we will try to make use of this normal
distribution curve - we will have some indices
60
00:05:07,240 --> 00:05:13,190
for the precision; also, we will go for the
weighted observations, what are the weighted
61
00:05:13,190 --> 00:05:20,580
observations and their measures of the precision,
then we will try to see - if the time permits,
62
00:05:20,580 --> 00:05:25,150
in this lecture - that how we can detect the
outliers, and what is here?
63
00:05:25,150 --> 00:05:27,020
- what happens in case of the error propagation.
64
00:05:27,020 --> 00:05:31,039
So this is all we are going to talk today.
65
00:05:31,040 --> 00:05:35,990
Now, what we will do, we will look a little
bit - of course, we have seen it before also
66
00:05:35,990 --> 00:05:39,430
- into the normal curve.
67
00:05:39,430 --> 00:05:50,190
The normal curve, as we saw yesterday - or
maybe in our last lecture - was like this.
68
00:05:50,190 --> 00:06:00,780
Any variable ‘x’ may have an arithmetic
mean which you can determine, and the probability
69
00:06:00,780 --> 00:06:04,340
of occurrence of this x is ‘Y’.
70
00:06:04,340 --> 00:06:15,919
So, this Y is written as ‘f(x)’, which
is given by 1 by sigma 2 pi - you can determine
71
00:06:15,919 --> 00:06:24,630
this sigma, standard deviation, we saw the
formula for that - e raised to power 1/2 of
72
00:06:24,630 --> 00:06:31,090
x minus X bar upon sigma square.
73
00:06:32,080 --> 00:06:40,099
Okay, our observations in surveying, our errors,
random errors - and this is true only if,
74
00:06:40,110 --> 00:06:44,940
from our observations, the blunders, mistakes
had been taken away; those observations which
75
00:06:44,940 --> 00:06:49,560
were blunders or mistakes, they have been
taken away, and also from our observations,
76
00:06:49,560 --> 00:06:52,660
the systematic errors have also been taken
away.
77
00:06:52,660 --> 00:06:57,639
Then only - if only the random errors are
there, then we can safely assume that our
78
00:06:57,639 --> 00:07:02,310
observations will follow this pattern or this
curve.
79
00:07:02,310 --> 00:07:09,380
Now, what we will do, we will try to see how
we can make use of this.
80
00:07:09,380 --> 00:07:22,229
Before that, what we will discuss, we will
discuss a ‘standard normal curve’ or ‘standard
81
00:07:22,229 --> 00:07:23,769
normal distribution’.
82
00:07:24,120 --> 00:07:25,460
Now, what is this?
83
00:07:25,940 --> 00:07:33,719
Generally, a normal distribution is written
as ‘mu sigma square’.
84
00:07:33,720 --> 00:07:41,199
The meaning is, the distribution for a variable
that we say’x’ is normal, and the normal
85
00:07:41,199 --> 00:07:48,259
distribution is given by two parameters: the
mean, which is the most probable value, and
86
00:07:48,259 --> 00:07:49,630
the standard deviation.
87
00:07:49,630 --> 00:07:54,180
So, these two, they define the normal distribution
curve.
88
00:07:54,180 --> 00:08:02,860
For a standard normal distribution, we write
it as 0 and 12 . So, you can understand, for
89
00:08:02,860 --> 00:08:13,250
a normal - a standard normal curve, if the
variable is ‘z’, the mean of the variable
90
00:08:13,250 --> 00:08:16,729
is 0 and the standard deviation is 1.
91
00:08:16,729 --> 00:08:24,721
Well, I can plot it also -
if you plo
92
00:08:24,721 --> 00:08:30,949
t it like this, the meaning is, this value
is 0, while the standard deviation, we will
93
00:08:30,949 --> 00:08:37,020
see in a while, is -1 and +1 here.
94
00:08:37,020 --> 00:08:41,968
We will see these sigma terms - how they are
coming here, what this point is - we will
95
00:08:41,969 --> 00:08:45,329
look into this in a while.
96
00:08:45,329 --> 00:08:51,449
So, for a standard normal curve, what we can
do, we can do slightly more - we can write
97
00:08:51,450 --> 00:08:55,740
it as ‘f(z)’, as in the figure here.
98
00:08:55,740 --> 00:09:10,690
The equation of that: 1 by sigma is 1, and
root 2 pi, e is to power -z2 by 2, because
99
00:09:10,690 --> 00:09:14,370
in this case, the mean is 0.
100
00:09:14,370 --> 00:09:18,430
So, it reduces to this.
101
00:09:18,430 --> 00:09:23,599
What we will do, we will make use of this
standard normal curve, because for a standard
102
00:09:23,600 --> 00:09:31,050
normal curve we can compute the probability
of occurrence of these values here.
103
00:09:31,050 --> 00:09:36,290
There are some standard tables - we will try
to make use of those tables in our computations,
104
00:09:36,290 --> 00:09:40,430
so this is why we need to discuss this.
105
00:09:40,430 --> 00:09:45,599
But before that, what I will do, I will write
a term here.
106
00:09:45,600 --> 00:09:55,610
For example, let us say there is a variable
X for which the mean is - sorry the mean is
107
00:09:55,610 --> 00:09:59,660
X bar, and the standard deviation is sigma.
108
00:09:59,660 --> 00:10:11,779
So, we can write this normal distribution
as X bar and sigma square, and this - we will
109
00:10:11,779 --> 00:10:20,999
show like this, where this X may vary from
minus infinite to plus infinite.
110
00:10:22,980 --> 00:10:32,620
Now, I write a term here: I write that probability,
that when my variable X is less than or equal
111
00:10:32,630 --> 00:10:35,860
to a value x.
112
00:10:35,860 --> 00:10:38,330
What is the meaning of this?
113
00:10:38,330 --> 00:10:45,440
I am looking if this small x, this value,
is somewhere here.
114
00:10:45,440 --> 00:10:46,440
What I am looking at?
115
00:10:46,440 --> 00:10:53,540
I am looking at: now, by writing this, what
is the probability of occurrence of my variable
116
00:10:53,540 --> 00:10:57,920
X so that it is smaller than the value x?
117
00:10:59,180 --> 00:11:08,579
Now, as you know, in this curve, at any point
here, this is the probability of occurrence
118
00:11:08,579 --> 00:11:10,380
of this particular value.
119
00:11:10,380 --> 00:11:11,600
What we are looking here for?
120
00:11:11,600 --> 00:11:15,660
We are looking here for all the values which
are less than and equal to x.
121
00:11:15,660 --> 00:11:19,650
So, what is that probability?
122
00:11:19,650 --> 00:11:24,519
That probability is basically the area of
the curve here.
123
00:11:24,519 --> 00:11:35,660
So, this area, we can also find as - I can
write it as, let us say, F(x) is integration
124
00:11:35,660 --> 00:11:47,719
of this curve from minus infinity to x, f(X)
is my this curve, into dx.
125
00:11:47,720 --> 00:11:55,209
So, what it will do, it will give me the area
of the curve in between these two.
126
00:11:55,209 --> 00:12:00,768
Now, let us look at this thing in a slightly
different way - what is this really?
127
00:12:03,320 --> 00:12:08,740
This is telling me the ‘cumulative probability’
- you know, the probability of occurrence
128
00:12:08,740 --> 00:12:15,240
of all the variables, all the values, from
minus infinite to x, is this value.
129
00:12:15,240 --> 00:12:20,649
So, this is actually the sum of all these
probabilities starting from here to there.
130
00:12:20,649 --> 00:12:26,470
It is starting from minus infinite to here
at x, because we are taking this area.
131
00:12:26,470 --> 00:12:34,870
So, we say this as ‘cumulative probability
distribution or density function’, because
132
00:12:36,160 --> 00:12:42,040
this is the probability density function,
and this is now - if f(X) is being plotted
133
00:12:42,040 --> 00:12:47,709
with x, that will become cumulative probability density function.
134
00:12:47,710 --> 00:12:58,399
Many times, this is also called as simply
‘distribution function’.
135
00:12:58,399 --> 00:13:03,730
Distribution function; this f(X).
136
00:13:05,140 --> 00:13:10,620
Now, we will try to plot these two together
– Okay!
137
00:13:10,620 --> 00:13:20,510
- I am plotting my normal curve first - here
is the normal curve from minus infinite to
138
00:13:20,510 --> 00:13:27,250
plus infinite for my variable X, this is the
probability of occurrence of any value here.
139
00:13:29,460 --> 00:13:44,040
Now, what I am doing, I am plotting, along
with this, in the same line here, my another
140
00:13:44,050 --> 00:13:53,810
curve, and this curve is called F(X) for - again, the value is X.
141
00:13:54,440 --> 00:14:01,860
Well, what we had done in our last slide,
we had computed the value of F(X) at x - this
142
00:14:03,500 --> 00:14:14,860
small x - and that was this area, and the
value was F(x), which we said the cumulative
143
00:14:14,860 --> 00:14:19,820
probability density function, or cumulative
probability density, or cumulative probability
144
00:14:19,820 --> 00:14:21,220
as such, we can say.
145
00:14:23,080 --> 00:14:30,500
Well, if I plot this x against this X here,
what we will get, we will get a curve like
146
00:14:32,500 --> 00:14:33,500
this.
147
00:14:33,500 --> 00:14:43,580
Now, you can very well tell me - you can guess it, now - that what will be the
value of this
148
00:14:43,580 --> 00:14:44,680
total one.
149
00:14:47,640 --> 00:14:48,939
What this total one is?
150
00:14:48,940 --> 00:14:53,240
This total one should be 1 - why?
151
00:14:53,240 --> 00:14:58,940
Because the probability of occurrence of all
these variables - all the values starting
152
00:14:58,940 --> 00:15:03,620
from minus infinite to plus infinite - all
of them, they will occur; what is the probability
153
00:15:03,620 --> 00:15:04,620
of that?
154
00:15:04,620 --> 00:15:05,890
The probability is 1.
155
00:15:05,890 --> 00:15:13,579
So, the cumulative probability will be 1,
while at this place, this particular value
156
00:15:13,579 --> 00:15:22,359
is F(x) at small x, and this value will be
1 minus F(x).
157
00:15:22,360 --> 00:15:30,470
So, for our standard normal curve, what we
are looking at?
158
00:15:30,470 --> 00:15:35,180
N - standard normal curve - 0, 12.
159
00:15:35,180 --> 00:15:43,138
These values are known; there are some standard
tables where all these values are known against
160
00:15:43,139 --> 00:15:44,390
the variable.
161
00:15:45,260 --> 00:15:47,819
We will make use of those in a while.
162
00:15:50,180 --> 00:15:59,120
Well, in order to make use of that, I am writing here now, again, for a variable -
let us say
163
00:15:59,120 --> 00:16:05,949
it is capital X - and its distribution is
normal distribution.
164
00:16:05,949 --> 00:16:11,109
The meaning is, from our observations, all
the systematic errors and blunders, they have
165
00:16:11,110 --> 00:16:14,560
been taken away - only we have got random
errors.
166
00:16:14,560 --> 00:16:16,900
This is why this distribution will be there.
167
00:16:18,040 --> 00:16:28,099
Well, what I try to find, I try to find now,
between two values x1 and x2 - so, these are
168
00:16:28,100 --> 00:16:40,899
two realisations x1 and x2 of my random variable
X. I am trying to find the probability that
169
00:16:40,899 --> 00:16:48,240
my variable is between x1 and x2.
170
00:16:48,240 --> 00:16:50,940
Now, what is the meaning of this, first?
171
00:16:50,940 --> 00:16:53,279
Let us understand the meaning.
172
00:16:53,279 --> 00:17:00,500
The meaning of this is, what is the probability
of occurrence of all the values which are
173
00:17:00,500 --> 00:17:07,569
between x1 and x2 , and this, as you can see
here, is actually this area.
174
00:17:07,569 --> 00:17:15,589
So, we need to integrate our curve starting
from this point to this point, and to do that,
175
00:17:15,589 --> 00:17:17,940
we should know the characteristics of the
curve.
176
00:17:17,940 --> 00:17:22,549
Now, what we will do, we will make use of
standard normal curve in order to find this
177
00:17:22,549 --> 00:17:23,549
value.
178
00:17:23,549 --> 00:17:35,480
Now, to do that, what we do, we find as - we
have written earlier also - Z. Z is the normalized
179
00:17:35,480 --> 00:17:36,480
variable.
180
00:17:36,480 --> 00:17:42,860
What we are doing, we are transforming this
variable x here and this transformation is
181
00:17:42,860 --> 00:17:53,699
X minus the variable X bar - X bar is the
mean of this variable - divided by sigma - the
182
00:17:53,700 --> 00:17:56,850
standard deviation of this variable.
183
00:17:56,850 --> 00:18:06,199
So, using this, we can transform our variable
capital X here into the standard variable
184
00:18:06,200 --> 00:18:09,400
or into the standard normal curve variable.
185
00:18:09,400 --> 00:18:20,840
Now, this particular variable Z has the normal
distribution, which is 0 and 12.
186
00:18:20,840 --> 00:18:22,409
Okay.
187
00:18:22,410 --> 00:18:23,929
Why we did it; what is the use of this?
188
00:18:23,929 --> 00:18:37,940
To see the use of this: the probability that
X -capital X - between x1 and x2 - the probability
189
00:18:37,940 --> 00:18:45,660
of occurrence of capital X between these two
variables is actually same as the probability
190
00:18:45,660 --> 00:18:55,100
that the normal standard variable occurs between
z1 and z2.
191
00:18:55,100 --> 00:18:57,990
Now, what these z1 and z2 are?
192
00:18:57,990 --> 00:19:05,039
Obviously, you can find z1 - again, the transformed
variable - and the z1 will be the transformed
193
00:19:05,039 --> 00:19:18,658
variable, x1 minus X bar by sigma and similarly
z2: x2 minus X bar by sigma.
194
00:19:19,300 --> 00:19:24,960
So, you can transform your values now - the
values x1 and x2 into z1 and z2.
195
00:19:25,960 --> 00:19:30,720
So, if you have done so, you would like to
determine the value of this.
196
00:19:31,360 --> 00:19:38,699
Because we are saying that we - it is possible to find this value from the standard
normal
197
00:19:38,700 --> 00:19:40,240
curve table.
198
00:19:40,660 --> 00:19:41,560
How to find this value?
199
00:19:43,220 --> 00:19:49,680
This value, we can write further, and to explain that, I am drawing, again, my
standard normal
200
00:19:49,690 --> 00:19:57,570
curve here - this is for my Z, where it is
0.
201
00:19:58,640 --> 00:20:07,059
Let us say z1 and z2 are here - this is my
z1 and this is my z2 corresponding to my x1
202
00:20:07,059 --> 00:20:09,080
and x2.
203
00:20:09,080 --> 00:20:14,770
Basically, what I am interested in, I am interested
in this area.
204
00:20:14,770 --> 00:20:22,510
The moment I write this is actually the area
here, the shaded area; the shaded part.
205
00:20:22,510 --> 00:20:23,970
How can I determine the shaded part?
206
00:20:23,970 --> 00:20:32,020
To determine the shaded part, well, let us
say, if I can determine this entire part which
207
00:20:32,020 --> 00:20:39,418
has the dots, what this will be?
208
00:20:39,419 --> 00:20:51,120
The one which has the dots will be the probability that my Z is less than or equal
to z2 -
209
00:20:52,200 --> 00:20:52,920
isn’t it?
210
00:20:53,440 --> 00:21:06,360
Then, if I can also determine the area here
- if I am drawing by these lines - can I determine
211
00:21:06,360 --> 00:21:07,360
this one?
212
00:21:07,360 --> 00:21:16,939
Let us say the area of this one is the probability
that my Z is less than or equal to z1.
213
00:21:16,940 --> 00:21:23,780
Now, having known this and this, if I can
differentiate, you can find the difference
214
00:21:23,780 --> 00:21:30,530
of these two - so this total area minus this
area here will give me the area under the
215
00:21:30,530 --> 00:21:31,750
curve here.
216
00:21:31,750 --> 00:21:38,740
So, by determining this and this, I should
be able to find this probability.
217
00:21:38,740 --> 00:21:45,410
Well, these are the values which you can get
from the standard normal curve, or from the
218
00:21:45,410 --> 00:21:49,280
standard normal curve table, for any value
of standard variable.
219
00:21:49,280 --> 00:21:54,980
Now, this is standard variable; these are
the standard tables of z1 and z2.
220
00:21:54,980 --> 00:22:00,320
You can find these two values, you can find
the differences, and by finding that, in fact,
221
00:22:00,320 --> 00:22:01,428
what you are finding?
222
00:22:01,429 --> 00:22:09,080
You are determining the probability of occurrence
of X in between x1 and x2.
223
00:22:09,080 --> 00:22:10,730
What we will do, we will make use of this
now.
224
00:22:10,730 --> 00:22:13,820
How it is useful to us; why we are talking
about it?
225
00:22:13,820 --> 00:22:15,549
So, we will make use of this.
226
00:22:15,549 --> 00:22:29,049
Well, to make use of this, what I will do,
I will draw again the normal curve for a variable
227
00:22:29,049 --> 00:22:36,710
X, and of course, this is the mean, and we
also know for this X, the corresponding sigma.
228
00:22:36,710 --> 00:22:43,799
Then we can say this is normally distributed
with the mean of x bar and sigma square.
229
00:22:43,799 --> 00:22:48,600
Actually, this is the conventional way, the
standard way, in which the normal distribution
230
00:22:48,600 --> 00:22:53,719
is represented by these two parameters; that
is why I am writing it every time.
231
00:22:53,720 --> 00:23:05,170
Then, our normal curve, the bell shaped curve,
which we say from minus infinite to plus infinite.
232
00:23:05,170 --> 00:23:11,392
Now, we want to make use of what we saw just
in the previous slide; we want to make use
233
00:23:11,392 --> 00:23:20,870
of the - can we - the idea is, can we really
find what is the probability of occurrence
234
00:23:20,870 --> 00:23:23,168
using the normal curve between two values?
235
00:23:23,169 --> 00:23:36,280
Now here, in this curve, this point and this
point are the points of inflexion - points
236
00:23:36,280 --> 00:23:47,299
of inflexion - and these points, they are
corresponding to the values of sigma - sigma
237
00:23:47,299 --> 00:23:50,980
on this side and sigma on this side.
238
00:23:50,980 --> 00:24:06,300
Well, can I really find the area of the curve
which is within mu - sorry, this X bar - plus
239
00:24:06,300 --> 00:24:07,590
minus sigma.
240
00:24:07,590 --> 00:24:10,299
What is the meaning of this?
241
00:24:10,299 --> 00:24:16,830
The meaning is - X bar, minus sigma on this
side, plus sigma on this side - so this area
242
00:24:16,830 --> 00:24:18,629
which is shaded by red colour here.
243
00:24:18,630 --> 00:24:23,340
So, what we can do, we can make use of the
standard normal curve and again, we can find
244
00:24:23,340 --> 00:24:30,959
this area, and this area comes out to be 68.3
percent.
245
00:24:32,330 --> 00:24:37,449
If we are considering the entire area within
the normal curve to be 100, then it is 68.3.
246
00:24:39,100 --> 00:24:46,899
Similarly, we can also get the area corresponding to plus-minus 2 sigma, because
our two sigma
247
00:24:46,909 --> 00:24:55,280
will be here somewhere, and that comes out
to be 95.5.
248
00:24:55,280 --> 00:25:08,899
We can also find this area corresponding to
3 sigma, which is 97.7 - sorry, 99.7.
249
00:25:08,900 --> 00:25:21,210
For plus-minus 3.29 sigma, the area is 99.9.
250
00:25:21,210 --> 00:25:25,549
What is the meaning of these things what I
am writing, why we need it, how we can make
251
00:25:25,549 --> 00:25:26,860
use of this?
252
00:25:26,860 --> 00:25:28,699
We should see this thing now.
253
00:25:28,700 --> 00:25:34,289
So, to see this thing, what I will do, I will
give you one little example here.
254
00:25:34,289 --> 00:25:41,440
Let us say we have a length; the true value
of the length is L - we do not know this;
255
00:25:41,440 --> 00:25:42,520
we cannot measure it.
256
00:25:42,520 --> 00:25:45,350
It is not possible to measure this.
257
00:25:45,350 --> 00:25:51,340
What we do in turn, we take several observations of this length L. For example, let
us say
258
00:25:51,340 --> 00:25:58,539
l1, l2, l3 and so on, up to ln.
259
00:25:58,539 --> 00:26:00,908
We take many observations.
260
00:26:00,909 --> 00:26:08,520
From these observations I can find ‘l bar’
- the mean of this - and, as well as, I can
261
00:26:08,520 --> 00:26:13,920
find the sigma, the standard deviation - you
know how to compute the standard deviation.
262
00:26:13,920 --> 00:26:24,010
So, once we know these two, also - the important
point: also, we are assuming here, the observations
263
00:26:24,010 --> 00:26:31,500
- the meaning is, from these observations,
the outliers or the mistakes have been taken
264
00:26:31,500 --> 00:26:32,500
away.
265
00:26:32,980 --> 00:26:39,500
Also, it means, from these observations, if
at all there was some error component - systematic
266
00:26:39,510 --> 00:26:41,850
error component - it had been also taken away.
267
00:26:42,300 --> 00:26:47,620
So, these observations have got only and only random errors within them.
268
00:26:48,580 --> 00:26:53,860
Now, in that case, because our assumption
is, if only random error is there, then our
269
00:26:53,860 --> 00:26:57,350
observations will follow the normal curve;
the normal distribution.
270
00:26:57,350 --> 00:27:05,629
Well, after finding this sigma and the mean,
how to report this observation?
271
00:27:05,630 --> 00:27:15,070
Reporting the observation -
how do we write it; what do we say about it?
272
00:27:15,070 --> 00:27:18,168
Something which we have found, how do we present
it?
273
00:27:18,169 --> 00:27:25,690
Well, I say ‘l bar plus-minus sigma’ is
my value - what does it convey?
274
00:27:25,690 --> 00:27:32,100
If you tell - if you measure this length here,
after measuring this length, you find the
275
00:27:32,100 --> 00:27:38,379
mean of that, you find the sigma of the observations
or standard deviation, and you say ‘Well,
276
00:27:38,380 --> 00:27:40,240
this is my mean, this is my standard deviation’.
277
00:27:40,240 --> 00:27:42,840
What is the meaning of this?
278
00:27:42,840 --> 00:27:47,529
What - you know, someone, a third person - what
should he understand out of this?
279
00:27:47,529 --> 00:27:51,700
What you as an observer, you as a surveyor
should understand out of this?
280
00:27:51,700 --> 00:28:02,950
The meaning is - because we knew already that
68.3 percent or 68.3 was the area within the
281
00:28:02,950 --> 00:28:11,370
curve - our standard normal curve - which
was falling within plus-minus sigma.
282
00:28:11,370 --> 00:28:12,719
What is the meaning now?
283
00:28:12,720 --> 00:28:24,240
The meaning is, we are sure now, we are sure
that out of 100 observations if we take, 68.3
284
00:28:24,240 --> 00:28:27,399
observations will be within this limit.
285
00:28:27,399 --> 00:28:34,340
This is what the normal curve says - the normal
curve gives you the probability of occurrence
286
00:28:34,340 --> 00:28:36,158
of any observation.
287
00:28:36,159 --> 00:28:42,679
And we have seen that within plus-minus sigma,
from the mean, the probability of occurrence
288
00:28:42,679 --> 00:28:48,620
of all the observations in that is 68.3 percent
- that, we have seen.
289
00:28:48,620 --> 00:28:55,850
So, if I am reporting my measurements like
this , so the meaning is, if I am taking 1000
290
00:28:55,850 --> 00:29:04,240
observations, then - if the observations have
got only the random error - 683 observations
291
00:29:04,240 --> 00:29:09,730
out of 1000 will be within plus-minus 1 sigma
of that.
292
00:29:09,730 --> 00:29:13,320
So, this gives you a kind of confidence about
your values.
293
00:29:13,320 --> 00:29:25,490
For example, let us say, I can say at 95.5
percent confidence - this is the confidence
294
00:29:25,490 --> 00:29:33,710
level now - at this level, I can say that
my observation is ‘l bar plus-minus 2 sigma’.
295
00:29:33,710 --> 00:29:35,970
What is the meaning?
296
00:29:35,970 --> 00:29:39,549
So many times, we are sure.
297
00:29:39,549 --> 00:29:50,200
For, example, if you are taking a little set
- for a population, l bar and sigma are the
298
00:29:50,200 --> 00:29:56,220
values which you know; the population is defined,
or the normal distribution curve of the population
299
00:29:56,220 --> 00:29:58,300
is defined, by these two.
300
00:29:58,300 --> 00:30:08,270
Then, if you take a little set from that population,
the mean of that set, of that sample, will
301
00:30:08,270 --> 00:30:16,879
have - in 95.5 percent times, it will be within
this, while this is for the population.
302
00:30:16,880 --> 00:30:22,360
So, for any set you take, this tells you about
confidence about your observation and this
303
00:30:22,360 --> 00:30:24,860
is how, you know, we report our observations.
304
00:30:24,860 --> 00:30:31,840
So mostly, you will find - any instrument
- if there is an instrument, you are going
305
00:30:31,840 --> 00:30:37,470
to buy an instrument, for the instrument,
let us say, a simple scale - well, what is
306
00:30:37,470 --> 00:30:39,630
the accuracy by which the scale can measure?
307
00:30:39,630 --> 00:30:45,469
So, the manufacturers, they will say, the
accuracy of my scale is, for example - anything
308
00:30:45,469 --> 00:30:52,030
- this is plus-minus 15 centimetre.
309
00:30:52,030 --> 00:30:57,470
You have got some instrument which can measure
the length, and the instrument says the accuracy
310
00:30:57,470 --> 00:30:59,870
of the measurement is plus-minus 15 centimetre.
311
00:30:59,870 --> 00:31:06,000
You should ask at that time: what is this
15 centimetre - is this 1 sigma, 2 sigma,
312
00:31:06,000 --> 00:31:09,540
3 sigma or at what sigma?
313
00:31:10,480 --> 00:31:17,440
This is important because depending on this, you will come to know what kind of
instrument
314
00:31:17,440 --> 00:31:19,760
is that; how good it is for your purpose.
315
00:31:20,940 --> 00:31:29,240
You know, at what - because if this accuracy
is given at 3 sigma; also, in the second instrument,
316
00:31:29,240 --> 00:31:36,820
it is given as 1 sigma - there are two instruments
instrument A or manufacturer A, and manufacturer
317
00:31:36,820 --> 00:31:43,549
B - they are building the same kind of instruments;
the manufacturer A, he says that his instrument
318
00:31:43,549 --> 00:31:50,100
can give at 1 sigma this accuracy, while he
is saying that at 3 sigma, you can get the
319
00:31:50,100 --> 00:31:51,279
same accuracy.
320
00:31:52,080 --> 00:31:54,260
What is the meaning; which one is more accurate?
321
00:31:54,950 --> 00:32:01,750
You will go for one instrument in which case this accuracy is available in more
number
322
00:32:01,750 --> 00:32:05,110
of observations, which is here.
323
00:32:06,260 --> 00:32:13,100
So, we should understand how we are making use of this standard or normal curve:
our
324
00:32:13,100 --> 00:32:20,580
concept, our observations, are - or our observations
are normally distributed; our errors are random
325
00:32:20,580 --> 00:32:21,580
errors.
326
00:32:21,580 --> 00:32:25,668
So, we should know all these things and then
only we can start making the judgement about
327
00:32:25,669 --> 00:32:30,360
the observations, about the observations which are reported, because this
particular observation
328
00:32:30,360 --> 00:32:33,260
or this particular error is reported.
329
00:32:33,260 --> 00:32:37,269
So, we can start making use of this - so,
this is important.
330
00:32:37,269 --> 00:32:42,830
Now what we will do, we will continue our
discussion in the same line.
331
00:32:42,830 --> 00:32:53,799
We saw in our last video lecture a term called
‘precision’.
332
00:32:53,799 --> 00:33:00,520
Now, we have seen so far that what is the
meaning of the precision, and I gave you one
333
00:33:00,520 --> 00:33:06,769
example where you are firing bullets in the
bulls-eye.
334
00:33:06,769 --> 00:33:09,100
What you are doing in that case?
335
00:33:09,100 --> 00:33:17,000
Let us say, there is a fireman and he is very
good; he fires all his bullets very closely,
336
00:33:17,000 --> 00:33:18,150
here.
337
00:33:18,150 --> 00:33:25,000
His observations are, or his firing is, not
accurate - obviously, because it is away from
338
00:33:25,000 --> 00:33:26,910
the bulls eye.
339
00:33:26,910 --> 00:33:28,490
But why is it happening?
340
00:33:28,490 --> 00:33:34,390
It is happening because of some forces which
are working there; some systematic errors,
341
00:33:34,390 --> 00:33:38,919
and because of those systematic errors all
the observations are clubbed here.
342
00:33:38,919 --> 00:33:45,740
Well, there is another person who is not a
very good fireman, and he also fires.
343
00:33:45,740 --> 00:33:54,660
Let us say, in his case, all his observations,
all his bullets, are this way.
344
00:33:56,040 --> 00:33:59,879
This is also not accurate because it is also
away from the bulls-eye.
345
00:33:59,890 --> 00:34:07,320
Again, in his case, some systematic error
is taking place, but if you look at these
346
00:34:07,320 --> 00:34:17,469
two, you will appreciate this person A for
his very, very precise firing than the person
347
00:34:17,469 --> 00:34:24,350
B. Just by looking at this diagram, we are
able to say, well, the person A is a better
348
00:34:24,350 --> 00:34:31,739
fireman than the person B. Now, how; what
made you think that?
349
00:34:31,739 --> 00:34:35,149
Because somewhere, you are talking of precision.
350
00:34:35,149 --> 00:34:45,980
That precision, over here, in these bullets,
is shown by the spread; the spread of bullets
351
00:34:45,980 --> 00:34:49,690
- are they spread widely or in very narrow
area.
352
00:34:49,690 --> 00:34:55,060
So this spread is an indicator of the precision.
353
00:34:55,060 --> 00:35:02,259
Now, what we do, we will try to take the same
concept into the surveying observations.
354
00:35:02,260 --> 00:35:05,150
Let us say there is a length.
355
00:35:05,150 --> 00:35:09,059
‘L’ is the true length, which we do not
know
356
00:35:09,059 --> 00:35:17,150
. There are two observers - A and B - both
of them, they measure this length.
357
00:35:17,150 --> 00:35:28,109
A also comes here, and he takes several observations,
for example, he takes l1, l2, l3, up to ln,
358
00:35:28,109 --> 00:35:35,750
and the B, B also takes several observations,
for example, let us say l1, l2 - I am writing
359
00:35:35,750 --> 00:35:44,180
them by red in order to differentiate - l3,
ln - or let us say lm- different number of
360
00:35:44,180 --> 00:35:45,700
observations - whatever.
361
00:35:45,700 --> 00:35:55,500
Now, just by looking at these observations,
can you speak about the precision - who is
362
00:35:55,500 --> 00:35:57,390
more precise out of these two?
363
00:35:57,390 --> 00:36:09,640
Well, what we do in this case, we will make
use of, again, our normal curve.
364
00:36:09,640 --> 00:36:14,650
As in this case, even if you eliminate the
systematic errors from your observations,
365
00:36:14,650 --> 00:36:20,869
all your observations shall shift here, and
they will still form a close group.
366
00:36:20,869 --> 00:36:27,760
Here also, if you eliminate the systematic
errors all the observations will shift somewhere
367
00:36:27,760 --> 00:36:32,240
here, and again they will form a widely spread
group.
368
00:36:32,240 --> 00:36:35,089
So, what we are doing?
369
00:36:35,089 --> 00:36:39,940
Maybe from these observations also, we have
taken away the systematic error component
370
00:36:39,940 --> 00:36:41,270
- it has been taken away.
371
00:36:41,270 --> 00:36:47,900
So, in your observations, only the random
error is there now, and this is why I am saying
372
00:36:47,900 --> 00:36:53,390
that now, henceforth, because it is only random
errors, we can make use of the normal curve.
373
00:36:53,390 --> 00:36:56,700
So, we will make use of the normal curve.
374
00:36:56,700 --> 00:37:04,939
Well, for this first set of observations,
I find the ‘l bar’ and the sigma - l
375
00:37:06,680 --> 00:37:12,279
uet's say, I should write it as ‘sigma a’
for this set of observations.
376
00:37:12,290 --> 00:37:21,700
B - from the B, I find again ‘l bar’ and
I find ‘sigma b’ - you can compute these.
377
00:37:21,700 --> 00:37:26,910
Of course, we are assuming in our observations that there is no systematic error
component,
378
00:37:26,910 --> 00:37:31,390
there is no outlier, or there is no any mistake
or blunder.
379
00:37:31,390 --> 00:37:40,049
Well, how to make use of these two in order
to see which observation is more precise?
380
00:37:40,050 --> 00:37:42,890
What I will do, I will try to plot it now.
381
00:37:42,890 --> 00:37:52,470
The plot of these observations may look like
- for B, let us say for B, the plot looks
382
00:37:52,470 --> 00:38:02,350
like this , and for A the plot looks like
- it is quite possible.
383
00:38:02,350 --> 00:38:06,460
Now, what is the meaning of this?
384
00:38:06,460 --> 00:38:18,060
The meaning of this is, for B, the standard
deviation will be the point of inflexion somewhere
385
00:38:18,060 --> 00:38:25,400
here, and that is the value of the standard
deviation, which we have already computed.
386
00:38:25,400 --> 00:38:33,840
This total is ‘2 sigma b’ - we have already
computed sigma B from the observations.
387
00:38:33,840 --> 00:38:47,400
For observer A, the point of intersection
should be somewhere here, and we have already
388
00:38:47,400 --> 00:38:54,600
computed ‘sigma a’, so this is ‘2 sigma
a’.
389
00:38:54,600 --> 00:39:06,740
Now, look at these two curves - do they say
anything about precision; are there any kind
390
00:39:06,740 --> 00:39:08,439
of indicators of the precision?
391
00:39:08,440 --> 00:39:17,540
Just look at that what we saw earlier: we
said, within plus-minus sigma, there are 68.3
392
00:39:17,540 --> 00:39:28,200
percent observations . So now, we are trying
to relate these two curves with the precision.
393
00:39:28,200 --> 00:39:30,629
This is the stick.
394
00:39:30,630 --> 00:39:38,470
We are talking that, within plus-minus sigma,
68.3 percent points or observations will be
395
00:39:38,470 --> 00:39:40,899
within plus-minus sigma.
396
00:39:40,900 --> 00:39:44,339
What is the meaning of that?
397
00:39:44,339 --> 00:39:50,960
The meaning is, here, in this case, the deviation
of the observations from the mean - how many
398
00:39:50,960 --> 00:39:51,960
observations?
399
00:39:51,960 --> 00:39:57,020
- 68.3 percent observations is only within
this limit.
400
00:39:57,020 --> 00:40:04,560
So, the deviation is less - the meaning is,
observations are closer; they are nearer to
401
00:40:04,560 --> 00:40:05,560
each other.
402
00:40:05,560 --> 00:40:11,920
While in the case of B, in the case of B,
the deviation of observations from the mean
403
00:40:11,920 --> 00:40:19,811
- because that is our sigma - 68.3 percent
observations are spread in this area, in a
404
00:40:19,811 --> 00:40:20,811
wider area.
405
00:40:20,811 --> 00:40:25,180
So, the deviation of the observations from
the mean is more here.
406
00:40:25,180 --> 00:40:31,350
This is something like, you know, a big spread,
while this particular one is like a smaller
407
00:40:31,350 --> 00:40:32,350
spread.
408
00:40:32,350 --> 00:40:34,078
So, what we are doing?
409
00:40:34,079 --> 00:40:41,010
We are relating our standard deviation with
the precision - standard deviation is an indicator
410
00:40:41,010 --> 00:40:42,310
of the precision.
411
00:40:42,310 --> 00:40:47,170
If you can find the standard deviation - so
here, in this case, the standard deviation
412
00:40:47,170 --> 00:40:57,510
of A is smaller than standard deviation of
B. So, we can say these observations, the
413
00:40:57,510 --> 00:41:02,839
A observations, are more precise than the
observations B, which is very obvious from
414
00:41:02,839 --> 00:41:03,839
here.
415
00:41:03,839 --> 00:41:06,000
Now, there is one question.
416
00:41:06,000 --> 00:41:13,420
The question is, for example, here only, in
this slide, both of them, they find for this
417
00:41:13,420 --> 00:41:15,490
length some mean values.
418
00:41:15,490 --> 00:41:21,589
Let us say, I am writing the this mean value
here as la and lb.
419
00:41:23,140 --> 00:41:29,220
Both of them, they have found some standard deviation also, and by the way,
incidentally,
420
00:41:29,220 --> 00:41:32,819
the standard deviations are equal.
421
00:41:35,960 --> 00:41:37,520
Now what is the end of this?
422
00:41:37,530 --> 00:41:42,910
Two observers, they went to the field, they
measured one length - one has done the observations
423
00:41:42,910 --> 00:41:49,490
n number of times, the other has done the
observations m number of times - and both
424
00:41:49,490 --> 00:41:55,580
of them, they come to the laboratory, they
present their results, they have one mean,
425
00:41:55,580 --> 00:41:59,830
one standard deviation; another mean, another
standard deviation.
426
00:41:59,830 --> 00:42:02,910
Incidentally, both the standard deviations
are same.
427
00:42:05,620 --> 00:42:12,799
Now, out of these two means, which one is
a better estimate of this length?
428
00:42:12,800 --> 00:42:19,060
The standard deviations are coming to be same;
can you say that the mean which is found from
429
00:42:19,060 --> 00:42:21,990
these observations is equally precise?
430
00:42:21,990 --> 00:42:29,959
Are the means same - are the means, you know,
with the same confidence, can you say so?
431
00:42:29,960 --> 00:42:35,670
The one thing which is different here - that
is about the number of observations.
432
00:42:35,670 --> 00:42:42,640
If we are taking only 10 observations in this
n, and we are coming with these two; in another
433
00:42:42,640 --> 00:42:48,650
case - let us say in case of B - we are taking
100 observations, and we are coming again
434
00:42:48,650 --> 00:42:50,530
with this.
435
00:42:50,530 --> 00:42:56,700
Are these two means equally confident; can
you be equally confident about these two means?
436
00:42:56,700 --> 00:43:01,180
So, in order to solve this problem - because
this really leads to the kind of confusion
437
00:43:01,180 --> 00:43:09,730
-there is a term which is called ‘standard
error’.
438
00:43:09,730 --> 00:43:19,010
What we do, we find the standard error as
the standard deviation divided by the root
439
00:43:19,010 --> 00:43:22,380
of number of observation.
440
00:43:22,380 --> 00:43:23,790
What we have done now, in this case?
441
00:43:23,790 --> 00:43:33,829
If, let us say, for the length which we measured,
we measured it n1 times and n2 times, while
442
00:43:33,829 --> 00:43:43,280
n1 is more than n2, and in both the cases
- for case A, case B - we had the same sigma.
443
00:43:43,280 --> 00:43:54,370
So, what we get, we get sigma by root of n1
and sigma by root of n2 . Because n1 is more
444
00:43:54,370 --> 00:44:03,040
than n2 - so, this is more, this value will
be less - the standard error in case of A,
445
00:44:03,040 --> 00:44:10,599
in case of this, is less, because of more
number of observations, though the standard
446
00:44:10,599 --> 00:44:11,599
deviation is same.
447
00:44:11,599 --> 00:44:13,430
What is the meaning of this?
448
00:44:13,430 --> 00:44:22,410
The meaning of this is, the mean which is
associated with this observation set has more
449
00:44:22,410 --> 00:44:28,319
precision - you are more confident about that
mean than the mean which is coming from here.
450
00:44:28,320 --> 00:44:35,700
So, the standard error is a way in which - because
in case of the standard deviation, it does
451
00:44:35,700 --> 00:44:40,720
not depend, you know, it does not take into
account the number of observations.
452
00:44:40,720 --> 00:44:49,180
Two sets of observations may have same standard
deviation, but that is wrong; the standard
453
00:44:49,180 --> 00:44:55,029
deviation - yes, it may be same, but when
we start talking about our mean - the value
454
00:44:55,030 --> 00:45:01,590
which you are finding from it; our result
- then we should not rely upon standard deviation
455
00:45:01,590 --> 00:45:02,590
alone.
456
00:45:02,590 --> 00:45:09,290
We should talk about standard error, because
here, we are considering also the number of
457
00:45:09,290 --> 00:45:15,319
observations which you have taken there in
the field, and the number of observations
458
00:45:15,319 --> 00:45:18,180
which are resulting in this standard deviation.
459
00:45:18,180 --> 00:45:23,759
So, this is important, and we should keep
this in mind whenever we are reporting; whenever
460
00:45:23,760 --> 00:45:25,980
we are judging the observations.
461
00:45:25,980 --> 00:45:43,430
Now, after this, we will start talking about,
now - weighted observation.
462
00:45:43,430 --> 00:45:49,589
What is the meaning of this?
463
00:45:49,589 --> 00:45:52,660
I will pose one question first.
464
00:45:52,660 --> 00:46:08,890
The question is, let us say, for a line there
are 5 persons: A, B, C and D and E. All they
465
00:46:08,890 --> 00:46:13,069
have done the measurements of this line.
466
00:46:13,069 --> 00:46:19,119
What they did, they went to the field with
a certain tape and then measured this length,
467
00:46:19,119 --> 00:46:28,310
you know, maybe on 5 different days, and they
come out to the lab with 5 sets of observations.
468
00:46:28,310 --> 00:46:36,049
Are all these observations of same reliability?
- that is the question.
469
00:46:36,050 --> 00:46:42,310
Let us say, right now, these 5 persons, they
go to the field and measure 5 lengths - la,
470
00:46:42,310 --> 00:46:45,940
lb, lc, ld and le.
471
00:46:45,940 --> 00:46:54,660
Now, question number one - question number
one, will all these - la, lb, lc, ld and le
472
00:46:54,660 --> 00:47:00,230
- will be the same, or will be the different?
473
00:47:00,230 --> 00:47:01,230
- question number one.
474
00:47:01,230 --> 00:47:07,130
Now, you should be able to answer this question
where the five individuals who are going to
475
00:47:07,130 --> 00:47:14,069
the field, with maybe the same instrument
- but will all these be same?
476
00:47:14,069 --> 00:47:20,150
No, they cannot be same, because there are
many varying factors.
477
00:47:20,150 --> 00:47:24,630
The person A, he looks at the instrument in
a slightly, you know, careless way; he does
478
00:47:24,630 --> 00:47:28,870
not look exactly the way the instrument should
be read.
479
00:47:28,870 --> 00:47:32,540
The person B - maybe he is committing some
blunder there.
480
00:47:33,160 --> 00:47:38,680
While person C is working, the temperature
is more; there is some error in this instrument
481
00:47:38,680 --> 00:47:41,118
- it may be the systematic error.
482
00:47:41,780 --> 00:47:46,200
Even if the way you put the instrument there
in the ground in order to measure the length,
483
00:47:46,210 --> 00:47:53,450
slightly some changes are there, what we saw
earlier, we can never measure the true values.
484
00:47:53,450 --> 00:47:57,368
Only if you can measure the true values - then
only all these observations would be same.
485
00:47:57,369 --> 00:48:02,720
So, all these observations will have to be
different, because the amount of the random
486
00:48:02,720 --> 00:48:08,328
error or systematic error which is there in
this observation is different.
487
00:48:08,329 --> 00:48:15,620
Even if you eliminate your systematic error,
still, these observations will be different,
488
00:48:15,620 --> 00:48:19,859
because the amount of the random error which
is there in the observations will be different.
489
00:48:19,859 --> 00:48:26,170
Well, so you have got five sets of observations;
five measurements of this length.
490
00:48:26,170 --> 00:48:33,660
Now, we want to take a mean; you want to find
the mean of this using these five.
491
00:48:33,660 --> 00:48:39,899
Should we, straight away from this ‘sigma
l by 5’?
492
00:48:39,900 --> 00:48:42,950
No, we should not do that.
493
00:48:42,950 --> 00:48:44,970
Why we should not do that?
494
00:48:44,970 --> 00:48:49,520
Because if you are doing it, we are considering
each and every observation to be having the
495
00:48:49,520 --> 00:48:53,980
same weight; same reliability.
496
00:48:53,980 --> 00:49:03,119
You knew the person A is very sincere, while
you also knew that person E is not that sincere.
497
00:49:03,119 --> 00:49:09,050
You know, when person C was working in the
field, the weather was very bad; the working
498
00:49:09,050 --> 00:49:11,700
conditions were very, very poor.
499
00:49:11,700 --> 00:49:17,899
If that is so, each of these observations
will be not equally reliable; they will have
500
00:49:17,900 --> 00:49:18,970
different reliabilities.
501
00:49:18,970 --> 00:49:24,930
So, this reliability of observations, we say
as ‘weight’ - the weight of the observation.
502
00:49:24,930 --> 00:49:31,890
An observation which is more reliable is having
higher weight; an observation which is having
503
00:49:31,890 --> 00:49:33,779
less reliability will have less weight.
504
00:49:33,780 --> 00:49:41,080
So, along with these observations, what we
do, we also write - we also make - a table.
505
00:49:44,240 --> 00:49:57,640
Now, along with these lengths, we also write
wa, wb, wc, wd, we, and these are the weights
506
00:49:57,640 --> 00:49:59,290
of these observations.
507
00:50:01,100 --> 00:50:05,720
Well, if our observations are weighted, how
can we proceed further?
508
00:50:05,720 --> 00:50:10,730
Now, each and every observation is having
a different weight.
509
00:50:10,730 --> 00:50:20,000
Number one thing - that is, how to decide
or give weights.
510
00:50:20,000 --> 00:50:26,210
How do you want to give the weight?
511
00:50:27,340 --> 00:50:28,460
There should be some methods.
512
00:50:28,460 --> 00:50:37,579
Well, in the methods, number one is: by your
‘personal judgement’.
513
00:50:37,579 --> 00:50:42,390
What is the meaning of that?
514
00:50:42,390 --> 00:50:45,069
Well, here, you need the personal judgement.
515
00:50:45,070 --> 00:50:51,010
I said A - the person A - is a very, very
reliable person; he is very sincere, he cannot
516
00:50:51,010 --> 00:50:53,619
commit mistake - I give him higher weight.
517
00:50:53,619 --> 00:50:58,490
Person D - he is not that good in working;
I give him less weight.
518
00:50:58,490 --> 00:51:04,189
Person C, when he was working in the field,
the weather conditions were very, very poor
519
00:51:04,190 --> 00:51:05,760
- I give him also less weight.
520
00:51:05,760 --> 00:51:10,280
Person B is - he is all right, so I give him,
you know, weight.
521
00:51:10,280 --> 00:51:14,540
It is just like that, the personal judgement
- you make the personal judgement and you
522
00:51:14,540 --> 00:51:16,230
assign the weight.
523
00:51:16,230 --> 00:51:17,230
Can there be some other method?
524
00:51:17,230 --> 00:51:29,099
Well, the other method is ‘number of observations’.
525
00:51:29,099 --> 00:51:32,140
What is the meaning of this number of observations?
526
00:51:32,140 --> 00:51:47,118
Well, here only, if I extend my this table
further, now, this length was measured by
527
00:51:47,119 --> 00:52:02,309
these observers 5 times, 10 times, 20 times,
40 times, 3 times . We assume - let us say
528
00:52:02,309 --> 00:52:10,390
it is given to you, all these observers A,
B, C and D are equally sincere, and they have
529
00:52:10,390 --> 00:52:15,730
done the observations of the same line different
number of times.
530
00:52:15,730 --> 00:52:18,830
Obviously, which observation has got more
weight?
531
00:52:20,040 --> 00:52:26,940
This observation, having 40 number of observations,
because a single thing we have measured 40
532
00:52:26,950 --> 00:52:28,029
number of times.
533
00:52:28,030 --> 00:52:31,440
So, we can give definitely give more weight
to that.
534
00:52:31,440 --> 00:52:39,390
So, the second method of giving weight is
proportional to the number of observations.
535
00:52:39,390 --> 00:52:40,950
Can there be any better method?
536
00:52:40,950 --> 00:52:54,740
The third one - we give this weight as by
the ‘variance’, or we can say, the precision.
537
00:52:54,740 --> 00:53:03,819
We know - for example, now, I will draw a
diagram here: the same length is being measured
538
00:53:03,819 --> 00:53:12,839
now by two observers A and B. A has done,
for example, let us say 100 measurements.
539
00:53:12,839 --> 00:53:20,328
B has also done 100 measurements for this
length L. Both of them - A and B - are equally
540
00:53:20,329 --> 00:53:24,599
sincere; the weather conditions when they
were working in the field are almost same
541
00:53:24,599 --> 00:53:29,130
- if that is the case, whom should we give
more weight?
542
00:53:29,130 --> 00:53:35,050
So what we do, for this 100 measurements and
for this 100 measurements, we can find the
543
00:53:35,050 --> 00:53:43,670
sigma, and the square of sigma is the variance.
544
00:53:43,670 --> 00:53:51,410
Now, why we are giving the weight as per the
sigma; what is the reason?
545
00:53:51,410 --> 00:53:53,841
Because it gives us the closest number of
observations.
546
00:53:53,841 --> 00:54:01,460
The observer who is more sincere, who is working
very well, the errors - the sources of the
547
00:54:01,460 --> 00:54:06,000
errors in the field are less - the chances
are, all the observations will be very near
548
00:54:06,000 --> 00:54:08,540
to each other; very, very close to each other.
549
00:54:08,540 --> 00:54:14,160
So, if the observations are closer, you want
to give them more weight.
550
00:54:14,160 --> 00:54:21,180
So, the weight is given now by variance.
551
00:54:21,180 --> 00:54:30,140
Inversely proportional to the variance - less
is the variance, less is spread, more is the
552
00:54:30,140 --> 00:54:31,520
weight - obviously.
553
00:54:31,520 --> 00:54:37,920
So, these are the methods by which we can
give weight to our observations.
554
00:54:37,920 --> 00:54:59,230
Now, what we will do, we will look for measures
of precision for weighted data or observations.
555
00:54:59,230 --> 00:55:04,360
Now, in case of the weighted data, what this
weighted data is?
556
00:55:04,360 --> 00:55:12,309
You have a length L measured as l1, l2, l3,
ln.
557
00:55:12,309 --> 00:55:22,050
Then, you have some weights: w1, w2, w3 and
wn to each of these observations.
558
00:55:22,050 --> 00:55:29,410
Well, how can you find - number one thing
- the weighted mean?
559
00:55:29,410 --> 00:55:36,259
Weighted mean - because, we will need the
weighted mean very often whenever we are working
560
00:55:36,260 --> 00:55:37,260
with the data.
561
00:55:37,260 --> 00:55:45,650
The weighted mean of this observation is ‘sigma
li’ - where ‘i’ is 1 to n - divided
562
00:55:45,650 --> 00:55:51,940
by - you can guess it very quickly now - that
is, ‘sigma wi’.
563
00:55:51,940 --> 00:55:57,069
That gives you the weighted mean; the weighted
arithmetic mean.
564
00:55:57,069 --> 00:56:01,329
Now, how about the measure of precision?
565
00:56:01,329 --> 00:56:11,810
The measure of precision - number one - we
say, ‘standard deviation for weighted observation’
566
00:56:11,810 --> 00:56:18,359
or weighted data - so the same thing.
567
00:56:18,359 --> 00:56:26,390
We just put a little ‘w’ here in order
to indicate that this is the weighted.
568
00:56:26,390 --> 00:56:37,071
This is, as we seen before, also ‘sigma
wi’, ‘xi’ minus ‘x bar square’ - of
569
00:56:37,071 --> 00:56:46,049
course, this xi and x bar are the mean and
the observation, and wi is the weight of the
570
00:56:46,049 --> 00:56:54,799
observation - divided by n minus 1 . So, this
gives you the standard deviation of the weighted
571
00:56:54,799 --> 00:56:56,200
data.
572
00:56:56,200 --> 00:57:05,649
Third: ‘corresponding to the standard errors’.
573
00:57:05,650 --> 00:57:17,690
The standard error - as we have written ‘SE’
earlier - weighted, is… divided by…‘i’.
574
00:57:17,690 --> 00:57:26,150
So, all these are equivalent, in terms of
their physical meaning, to what we have already
575
00:57:26,150 --> 00:57:27,150
discussed.
576
00:57:27,150 --> 00:57:31,270
We have discussed the standard deviation - what
it means; standard error - what it means,
577
00:57:31,270 --> 00:57:36,560
where we use it; so all of these measures
are also similar to that.
578
00:57:36,560 --> 00:57:44,600
So, what we have done, we have discussed about
the errors, now - the normal distribution;
579
00:57:44,600 --> 00:57:50,200
how we make use of that, standard normal curve;
how we make use of that, if we are taking
580
00:57:50,200 --> 00:57:55,020
the observations how to report it, what is
the meaning of plus-minus sigma.
581
00:57:55,020 --> 00:57:59,430
Then, we saw the precision; how we can give
the precision.
582
00:57:59,430 --> 00:58:02,940
This is all what we have discussed in this
video lecture.
583
00:58:02,940 --> 00:58:08,150
Now, in our next video lecture, again, we
will continue with the errors in the measurement.
584
00:58:08,150 --> 00:58:11,860
So, the sixth lecture of this module will
be again in the same.
585
00:58:11,860 --> 00:58:16,360
So we end this lecture today and now - thank
you very much.