![Page 1: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/1.jpg)
Chapter 5Estimating Parameters From
Observational Data
Instructor: Prof. Wilson Tang
CIVL 181 Modelling Systems with UncertaintiesCIVL 181 Modelling Systems with Uncertainties
![Page 2: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/2.jpg)
Estimating Parameters From Observation Data
REAL WORLD “POPULATION”(True Characteristics Unknown)
Th
eore
tica
l Mod
el
Sample {x1, x2, …, xn}
Sampling(Experimental Observations)
Real Line -∞ < x < ∞
With Distribution fX(x)
Random Variable X
Inference On fX(x)
fX(x)
22 Variance
xMean
s
22
1
1
1
xxn
s
xn
x
i
iStatistical
Estimation
Role of sampling in statistical inference
![Page 3: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/3.jpg)
Point Estimations of Parameters, e.g. , 2, , etc.
a) Method of moments: equate statistical moments (e.g. mean, variance, skewness etc.) of the model to those of the sample.
22ˆ ,ˆ ; ,: normalin e.g. sxNX
From Table 5.1 in pp 224 – 225
See e.g. 5.2 in p. 227
22
2
1 Xar
2
1expXE
,LN:X lognormalin
2
seXEV
x
![Page 4: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/4.jpg)
Common Distributions and their Parameters
![Page 5: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/5.jpg)
Common Distributions and their Parameters (Cont’d)
![Page 6: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/6.jpg)
b) Method of maximum likelihood:
Parameter = r.v. X with fx(x)
Definition: L() = fX(x1,) fX(x2,)fX(xn,), where x1, x2,xn are observed data
of estimationopt 0L
Physical interpolation – the value of such that the likelihood function is maximized
(i.e. likelihood of getting these data is maximized)
For practical purpose, the difference between the estimates obtained from these different methods would be small if sample size is sufficiently large.
![Page 7: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/7.jpg)
b)Method of maximum likelihood (Cont’d):
= 1
= 2
e-x
x
fX(x)
X1 X2
Given X1 = 2 more likely
Similarly, X2 = 1 more likely
Likelihood of depends on fX(xi) and the xi’s
ˆ0d
dL
eeL 21 xx
![Page 8: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/8.jpg)
? estimating is X is good How
X: ,
n21 X...XXn
1X
n
n...
n
1X...XXE
n
1XE n21
What would you expect the value of X to be?
As n Var(X)
Before collections of data, X1 is a r.v. = X
X is r.v.
sampling) random todue s.i. (Assume
1
...1 2
22212 nn
nXXXVar
nXVar n
![Page 9: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/9.jpg)
What is the distribution of X?
n1
n2
n
X
n1 > n2
normal approx. is X large isn but normalnot X If
normal is X normal is X If
![Page 10: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/10.jpg)
Confidence interval of
We would like to establish P(? < < ?) = 0.95
n,Nx known assuming E4.1 see 0,1N is
n
xy
0.95
k0.025
= 1.96-1.96
0.025
n
x
95.096.1n
x96.1P
n96.1xx
n96.1
n96.1xx
n96.1
n96.1x
n96.1x Similarly
95.0n
96.1xn
96.1xP
![Page 11: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/11.jpg)
Confidence interval of (Cont’d)
0.65 assume 5.6;x 25;n data From
95.0855.5345.5P
Not a r.v. confidence interval
2
-1k where
nkx
nkx short,In
1-
2
2-1
025.095.0
/2
1 –
k/2
![Page 12: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/12.jpg)
E5.5
Daily dissolved oxygen (DO)n = 30 observationss = 2.05 mg/l assume = x = 2.52 mg/l
Determine 99% confidence interval of
005.0201.099.01 58.2995.0
005.01k1
1005.0
1.56;3.49 30
2.052.582.52
nkx 005.099.0
25.3;76.1 Similarly 95.0
As confidence level interval <> n <>
![Page 13: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/13.jpg)
n
stx;
n
stx
1n,21n,21
1-nf parameter on with distributistudent t
S
-X ofon distributi theNeed
n
Confidence Interval of when is unknown
Small f
Large f N(0,1)
0
known
level confidence samefor interval observe T is 1-n
nS
X
![Page 14: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/14.jpg)
/2
p
t/2,f0
3.49 to1.56 wider than
3.55 to1.4930
05.2756.252.2
756.2tt
995.0p005.02%99
291n
05.2s,52.2x
Ex.5.6,
p383 toGo
99.0
29,005.01n,2
(for known case)
![Page 15: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/15.jpg)
E5.9
Traffic survey on speed of vehicles. Suppose we would like to determine the mean vehicle velocity to within 2 kph with 99 % confidence. How many vehicles should be observed?Assume = 3.58 from previous study
nkx
21
Scatter
21n2n
58.3k 005.0
2.58
What if not known, but sample std. dev. expected to s = 3.58 and desired to be with 2 ?
![Page 16: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/16.jpg)
E5.9 (Cont’d)
559.058.3
2
n
t
2n
58.3tthen
1n,005.0
1n,005.0
Compare withn 21for known25 n 559.0
25
2.797 ,25n
557.026
2.787 ,26n
503.030
2.756LS ,30n
![Page 17: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/17.jpg)
ns
nkx
ns
nkx
k
n
n
n
n
1,-1
1,-1
0.95
0.95
tx ; (
t-x ; )
1
n
-xP by writingstart
( 95.0k-xP loadfor
) 95.0k-xPstrength for
limit confidence sided-One
Lower confidence limit
Upper confidence limit
1 –
k
Not /2
known unknown
![Page 18: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/18.jpg)
n
std
n
stx
n
n
1,2
1,2
ii
x - d
x- d
-
distance about true ddistributeNormally is distance measured 5.7 Fig.
A , D, e.g. quantities geometric of sEstimation
Theoryt Measuremen of Problem
distance.mean theofdeviation standard
distance. estimated theoferror standard
n
sD
n
dVar
2
d
s
d t,measuremenmean of dev. std. error Standard
- Similar to estimations of
![Page 19: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/19.jpg)
041.04
0817.0
n
s error std.
2.13 to87.14
0817.0182.30.2
182.3 t 0.95 -1for
0.0817s
00667.03
11.01.0
1s
0.2
2.0 2.0, 1.9, 2.1,
times4 measured is distance a e.g.
95.0
0.025,3
22
2
2
n
dd
d
i
![Page 20: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/20.jpg)
What about an area?
B C
D
No. Mean Sample Var. D B C
9 4 4
60 m 70 30
0.81 0.64 0.32
6000307060
CBD A
A
A of estimate need
CBD
A ofon distributi normal assumeA oferror std. m 421764
m 1764
4
32.060
4
64.060
9
81.0100
CB
AVar
A oferror std. need
2
A
4
222
2222
222
C
C
B
B
D
D
n
SD
n
SD
n
S
CVarC
ABVar
B
ADVar
D
A
![Page 21: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/21.jpg)
0 s' allfor 42 Vs 6.46
2171.3 407.3 1764 44
5.0DD2 1764
B
A2 1764 AVar
? 0 s' sother' 0.5 C,B ifWhat
A
CB
CBCB
ss
C
A
2
025.095.0
m 6083.1 to9.5916
4296.16000
A
kAA
![Page 22: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/22.jpg)
In general,
r1r2
h
2
22
1 rrA tan1h
error std.ZVar
n
s ,
n
s ,DVar where
Z
DVar ZVar
tmeasuremenmean ,...,D Z Z estimateBest
value true- ..., valueTrue
j
j
i
i
D
D
j
D
Di
2
i
21
i
i
2
mean
21
i21
i
i
D
D
jiijjii
k
k
n
s
ZVark
D
Z
D
Z
D
Z
DD
Z
![Page 23: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/23.jpg)
Interval Estimation of 2
X Normalfor
1
2 SVar
SE
1
1S
42
22
22
n
XXn i
1,
2
12
2
1(
-1or 95.0 ? P
nc
sn
sample variance
2 statistics – confidence leveln – no. of sample
![Page 24: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/24.jpg)
E 5.13
DO data: n = 30, s2 = 4.2
229,05.0
95.02
1,
2
12
mg/l 89.617.7
4.229
2.4130
(
1(
c
c
sn
n
![Page 25: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/25.jpg)
Estimation of proportions
p̂Var n as
n
p̂-1p̂kp̂ p
n
p-1p p̂Var
pp̂E trialsof no.
successes of no. p̂
21
n
x
![Page 26: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/26.jpg)
10 out of 50 specimens do not have pass CBR requirement.
E 5.14
0.91 to0.69 50
0.8-10.81.960.8p
compacted wellembankment of proportion p 8.050
40p̂
95.0
![Page 27: Chapter 5 Estimating Parameters From Observational Data Instructor: Prof. Wilson Tang CIVL 181 Modelling Systems with Uncertainties](https://reader036.vdocument.in/reader036/viewer/2022062515/56649d2c5503460f94a01d99/html5/thumbnails/27.jpg)
Review on Chapter 5
sided 2or 1
unknown andknown
XVar
XE
n
1xˆ
intervals Confidence 2.
etc... ˆ ,ˆ ,ˆ estimatesPoint 1.
2
n
s
xi
n
ppk
c
s
n
ˆ1ˆp̂p ;
n
xp̂
1-n(
on intervals Confidence 5.
approx. seriesTaylor - A ,
problemst Measuremen 4.
size sample of No. 3.
21
1,
2
12