Environmental Engineering Reference
In-Depth Information
4.2.1 Independent observations
Let
d
1
,
d
2
, …, and
d
n
denote
n
observed data. Assuming these observations are statistically
independent, the likelihood of θ, denoted as
l
(θ|
d
), which is the chance to observe the data
d
= {
d
1
,
d
2
, …,
d
n
}, can be written as follows:
n
∏
1
l
(
θ
|
d
)
=
f
(
d
|
θ
)
(4.1)
i
i
=
n
d
|θ is the joint probability of observing
d
given θ, which is the product of
n
probability terms [
f
(
d
i
|θ),
i
= 1,
n
] since
d
1
,
d
2
, …, and
d
n
are statistically independent obser-
vations. Based on the maximum likelihood principle, the optimal values of θ can be obtained
by maximizing the above likelihood function, or equivalently, the logarithm of the likeli-
hood function is as follows:
where
i
f
(
)
i
=1
n
∑
1
L
(
θ
|
d
)
=
ln (
f
d
|
θ
)
(4.2)
i
i
=
The same optimal values of θ will be obtained regardless of whether the likelihood func-
tion
(
Equation 4.1
) o
r the logarithm of the likelihood function
(
Equation 4.2
)
is maximized.
However, maximizing the latter is more efficient computationally.
eXaMPLe 4.1
Table 4.1
lists the shear strength data of a clay core of a gravity dam. Suppose the friction
angle (φ) follows the normal distribution and we are interested in its mean and standard
deviation, which are denoted herein as μ and σ, respectively.
Let φ
1
, φ
2
, …, φ
n
denote
n
observed values of the friction angle φ, respectively. In this
example,
d
i
= φ
i
, and
d
= {
d
1
,
d
2
, …,
d
n
} = {φ
1
, φ
2
, …, φ
n
}. Based on the probability density
function of a normal distribution (e.g., Ang and Tang 2007), the chance to observe the
data
d
given θ, which is the likelihood of interest, can be written as follows:
Table 4.1
Test data from the Ankang Hydropower site
Test no.
c (kPa)
φ
(
°
)
Test no.
c (kPa)
φ
(
°
)
1
165
11.86
14
85
20.81
2
127
14.04
15
18
25.64
3
253
13.50
16
15
22.29
4
427
10.20
17
78
24.70
5
106
11.31
18
12
26.10
6
242
12.95
19
34
22.78
7
209
12.41
20
70
19.80
8
328
13.50
21
20
17.74
9
98
12.95
22
20
20.81
10
10
15.64
23
217
20.30
11
213
16.17
24
221
20.30
12
365
17.22
25
254
27.47
13 324 20.81
Source: Data from Tang, X.S. 2013.
Computers and Geotechnics
, 49, 264-278.
Search WWH ::
Custom Search