1 Choice of Distribution Choice of Distribution 1. Theoretical Basis e.g. CLT, Extreme value 2. Simplify calculations e.g. Normal or Log Normal 3. Based on data: - Histogram - Probability paper
Dec 20, 2015
11
Choice of DistributionChoice of Distribution
1. Theoretical Basise.g. CLT, Extreme value
2. Simplify calculationse.g. Normal or Log Normal
3. Based on data:- Histogram- Probability paper
22
(a)Arrange the data in ascending order
(b)Let
(c)Plot VS
(d)See if follows a straight line
Nm xxxx ,...,...,, 21
1
N
mPm
mx mP
Use of Probability PaperUse of Probability Paper
33
Arithmetic Plot Probability Paper
1
0 x
)(xFx
Pro
bab
ility
Sca
le
x
)(xFx
Concept of Probability PaperConcept of Probability Paper
44
ExampleExample
Data: 2.9, 3.5, 4, 2.5, 3.1 N=5
m
1 2.5 1/6= 0.1667
2 2.9 2/6= 0.3333
3 3.1 3/6= 0.5000
4 3.5 4/6= 0.6667
5 4 5/6= 0.8333
mx 1/ NmPm
55
Normal Probability paper
75.02.395.3
ˆ
2.3ˆ
5.084.0
5.0
xx
x
50% 84%
4
3
2
1
3.2
3.95
66
Shear strength of concreteShear strength of concrete
m Shear strength s
m/N+1 lnS m
Shear strength s
m/N+1
lnS
1 0.35 0.0714 -1.05 8 0.58 0.5714 -0.54
2 0.40 0.1429 -0.92 9 0.68 0.6429 -0.39
3 0.41 0.2143 -0.89 10 0.7 0.7143 -0.36
4 0.42 0.2857 -0.87 11 0.75 0.7857 -0.29
5 0.43 0.3571 -0.84 12 0.87 0.8571 -0.14
6 0.48 0.4286 -0.73 13 0.96 0.9286 -0.04
7 0.49 0.5000 -0.71
77
Normal Probability Paper
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
S
88
0
-0.2
-0.4
-0.6
-0.8
-1.0
-1.2
Normal Probability Paper
-0.58
-0.22
50% 84%
0.36 -0.58;
withlognormal is so
S
S
S
36.0
)58.0(22.0
58.0
ln
ln
lnS
99
Lognormal Probability Paper
S
1.0
0.7
0.5
0.3
58.056.0lnln mx
1010
• Even though the data points appear to fall on a straight line, but how good is it?
• Would it be accepted or rejected at a prescribed confidence level?
• If it appears to fit several probability models, which one is better?
Goodness of fit test of distribution• Chi-square test ( )
• Kolmogorov-Smirnov test (K-S)
2
1111
Procedures of Chi-Square test ( )Procedures of Chi-Square test ( )
1. Draw histogram2. Draw proposed distribution (frequency
diagram) normalized by no. of occurrence same area as histogram
3. Select appropriate intervals4. Determine = observed incidences per interval
= predicted incidences per interval
based on model
2
in
ie
1212
Procedures of Chi-Square test ( )Procedures of Chi-Square test ( )
5. Determine for each interval
6. Determine for all intervals
Note: Larger Z less fit
7. Compare Z with the standardized value
2
i
ii
e
en 2)(
k
i i
ii
e
enZ
2)(
fC ,1
level of confidence No. of parameters in
proposed distribution, estimated from data
f = k – 1 – m
1313
• Validity of method rely on (combine some intervals if necessary)
55 iek ,
8.Check:
If probability model substantiated with confidence level
Otherwise Model not substantiated
fCZ ,1 1
1414
1515
Example 6.7 –Cracking strength of concrete
= 10.73
= 7.97
f = 8 – 3 = 5
= 11.1
As both & <
Both models substantiated
(LN is better than N)
NZ
LNZ
5 ,95.0C
NZLNZ 5 ,95.0C
1616
Kolmogorov-Smirnov (K-S) TestKolmogorov-Smirnov (K-S) Test
)(xSn
n
kk
xx
xxxn
k
xx
1
0
1
1{Arrange the data in ascending order:
Nk xxxx ,...,...,, 21
Sample CDF
1717
• Compare of sample with CDF, of proposed model.
• Identify the largest discrepancy between the two curves.
• Compare with a standardized value
reject model
model substantiated
)(xSn)(xFX
maxD
maxD
DD
DD
max
max
1818
265.005.026 D