a brief workshop on measurement uncertaintyresource.npl.co.uk/docs/science_technology/ionising...a...
TRANSCRIPT
![Page 1: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/1.jpg)
A Brief Workshop on Measurement Uncertainty
John Keightley
National Physical Laboratory
RCUF, NPL25 June 2008
![Page 2: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/2.jpg)
Uncertainties : The Easy Solution
“If your experiment needs statistics, you should have done a better experiment”
Ernest Rutherford(1871 -1937)
![Page 3: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/3.jpg)
Uncertainties
• Principle
– There is some uncertainty about every measurement result … a “margin of doubt”.
• What confidence do WE have in our values?• What confidence do OTHERS have in our
values?• Can our values withstand objective scrutiny, or
be defended in a COURT OF LAW ?
![Page 4: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/4.jpg)
Measurement Uncertainty : Basic Concepts
• A measurement result is incomplete without a statement of the associated uncertainty
• When you know the uncertainty in a measurement, then you can judge its fitness for purpose
• Good measurement practice can reduce uncertainty
• Understanding measurement uncertainty is the first step to reducing it.
![Page 5: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/5.jpg)
Where do errors and uncertainties come from ?
• The measuring instrument : bias, drift, noise …• The condition being measured, which may not be stable• The measurement process itself• “Imported” uncertainties
– Calibration uncertainties (Better than not calibrating at all) !
• Operator skill• Environmental conditions (temperature, pressure,
humidity etc…)• Others …
![Page 6: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/6.jpg)
Literature
GPG 11GPG 11
A BeginnerA Beginner’’s Guide s Guide to Uncertainty of to Uncertainty of
MeasurementMeasurement
NPL 2001NPL 2001
![Page 7: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/7.jpg)
Literature
GPG 49GPG 49The Assessment of The Assessment of
Uncertainty in Uncertainty in Radiological Radiological
Calibration and Calibration and TestingTesting
NPL 2003NPL 2003
![Page 8: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/8.jpg)
![Page 9: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/9.jpg)
ISO Guide to Uncertainty in Measurement : “GUM”, 1995
Provides internationally agreed concepts, recommendations, procedures for uncertainty evaluation.
Underlying philosophyMeans, standard uncertaintiesDistributionsSensitivity coefficientsCombination of uncertaintiesUncertainty budgets
![Page 10: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/10.jpg)
Measurement Uncertainty
• From the “GUM”– “parameter associated with the result of a
measurement, that characterises the dispersion of the values that could reasonably be attributed to the measurand”.
• i.e.: a standard deviation or the half width of an interval having a stated level of confidence
• the measurement result is the “best estimate” of the value of the measurand, and all components of uncertainty contribute to the dispersion.
• Simplified definition– quantified doubt about the measurement result.
![Page 11: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/11.jpg)
GUM : Basic Concepts
• Identify quantities that influence the measurand• Develop a model to account for the inter-relations of the
input quantities that influence the measurand
Model
Y = f(X1,X2,...XN)X2
X1
XN
YInput
quantities
Measurand...
![Page 12: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/12.jpg)
GUM : Basic ConceptsKnowledge of each input quantity is incomplete, andmust be expressed as a probability density function (PDF)
The expectation of the PDF is taken as the best estimate of the value for that quantity
The standard deviation of the PDF is taken as the standard uncertainty associated with that estimate
The PDF is based on knowledge of the input quantity that may be inferred from repeated measurements (TYPE A) or scientific judgement based on all available information on the variability of that quantity (TYPE B).
![Page 13: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/13.jpg)
GUM : Basic Concepts
Y = f(X1,X2,...XN)x2,u(x2)
x1,u(x1)
xN,u(xN)
y,u(y)
f(X1)
X1
normal
F(X
2)
X2
rectangular
...
F(X N
)
XN
triangular
F(Y
)
Y
![Page 14: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/14.jpg)
Steps to Evaluate Measurement Uncertainty• Identify input quantities to measurement model• Estimate uncertainties on inputs
– (including probability distributions : form of the spread of results)• Consider correlations between input parameters• Establish sensitivity coefficients
– How the output changes with respect to changes in the inputs• Calculate standard uncertainty on output quantity for each
input quantity• Combine standard uncertainties
– Propagation of Errors– Central Limit Theorem (combined uncertainty distribution)
• Calculate effective degrees of freedom• Report expanded uncertainty at k =2
![Page 15: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/15.jpg)
Two ways to evaluate individual components: Type A evaluations -
uncertainty estimates using
statistics (usually from repeated readings) Type B evaluations -
uncertainty estimates from any
other information,
e.g. from• past experience of the measurements• from calibration certificates • manufacturer’s specifications • from calculations • from published information • and from common sense, experience, feel …
Not necessarily “random
versus systematic”
Evaluating Uncertainty
![Page 16: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/16.jpg)
Basic Statistics : repeat measurementsFrom repeated measurements you can:
• Calculate an average or mean• to get a better estimate of the “true” value
• Estimate the standard deviation• quantifies how different individual readings
typically are from the mean of the whole set
• Estimate the standard deviation of the mean (SDOM)• estimate of the standard deviation of the
distribution of means that would be obtained if the mean were measured many times.
• NB : “True” standard deviation can only be found from an infinite set of readings.
( )
( ) ( )( )
1
2
1
2
1
1
1
nii
nii
nii
xx
n
x xs
n
x xss xn nn
=
=
=
=
−=
−
−= =
−
∑
∑
∑
![Page 17: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/17.jpg)
Basic Statistics : PDFs( )
( ) ( )
( ) [ ]{ }( )( ) ( )
2
2 2 2
22
: .
: 2
2
X
X
X X
For a continuous random variable X, with probability density function f x
Expectation E X x f x dx 1st Ordinary Moment
Variance Var X E X nd Central Moment
E X
E X E X
For jointly distribu
μ
μ
μ μ
+∞
−∞= =
= −
= − +
= −
∫
( ) ( )
( ) ( )( )( )( ) ( ) ( )( ) ( ) ( )
: ,
X Y
X Y
Y X X Y
Y X X Y
ted random variables X and Y, with expectations E X and E Y
Covariance Cov X Y E X Y
E XY X Y
E XY E X E Y
E XY E X E Y
μ μ
μ μ
μ μ μ μ
μ μ μ μ
= =
⎡ ⎤= − −⎣ ⎦= − − +
= − − +
= −
![Page 18: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/18.jpg)
Probability Distributions• Describe the form of the spread of results
– Normal (Gaussian)• Most common
– Rectangular (Uniform)• Results evenly spread
– Triangular• Min, max and “most likely” known
– U-shaped• Most likely at ends of distribution
![Page 19: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/19.jpg)
Rectangular Distribution
f(x)
xa b(a+b)/2
1/(b-a)
( )1
0
a x bf x b a
otherwise
⎧ ≤ ≤⎪= −⎨⎪⎩
[ ] ( )( )( )
( )
[ ] ( ) ( )
( )
2 2 2
3 2 22 2
22 22 222
1 1. 2 2 2 2
1 1. 3 3
23 4 12
2 3
bb
a abb
a a
x
b a b ax b a a bE x mean x dxb a b a b a b a
x a ab bE x x dxb a b a
a ab b b aa ab bE x E x Variance
b aσ
+ −⎡ ⎤ − += = = = = =⎢ ⎥− − − −⎣ ⎦
⎡ ⎤ + +⎡ ⎤ = = =⎢ ⎥⎣ ⎦ − − ⎣ ⎦
+ + −+ +⎡ ⎤ − = = − =⎣ ⎦
−=
∫
∫
![Page 20: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/20.jpg)
Triangular Distribution
f(x)
xa b(a+b)/2
2/(b-a)
( )
( )( )
( )( )
2
2
42
42
0
a bx a a xb a
a bx b x bf xb a
otherwise
+⎧ − ≤ ≤⎪ −⎪⎪ − +⎪ − ≤ ≤= ⎨
−⎪⎪⎪⎪⎩
[ ]( )
( ) ( )
( )( ) ( )
[ ] ( ) ( )
( )
2
2
2
2 222 2 2
2
2
22 22 222
4 2
4 7 10 7 24
6 12 67 10 724 24 24
2 6
a bb
a ba
a bb
a ba
x
a bE x mean x x a dx x x b dxb a
a ab bE x x x a dx x x b dxb a
a ab b b aa ab bE x E x Variance
b aσ
+
+
+
+
⎡ ⎤+⎢ ⎥= = − + − =⎢ ⎥− ⎢ ⎥⎣ ⎦
⎡ ⎤+ +⎢ ⎥⎡ ⎤ = − + − =⎣ ⎦ ⎢ ⎥− ⎢ ⎥⎣ ⎦
+ + −+ +⎡ ⎤ − = = − =⎣ ⎦
−=
∫ ∫
∫ ∫
![Page 21: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/21.jpg)
U-Shaped Distribution
f(x)
xa b(a+b)/2
2/(b-a)
( )
( )( )( )
( )( )( )
2
2
2 22
2 22
0
a bx a b a xb a
a bx a b x bf xb a
otherwise
− +⎧ − + ≤ ≤⎪ −⎪⎪ +⎪ − + ≤ ≤= ⎨
−⎪⎪⎪⎪⎩
[ ]( )
( )( ) ( )( )
( )( )( ) ( )( )
[ ] ( ) ( )
( )
2
2
2
2 222 2 2
2
2
22 22 222
2 2 2 2
2 3 2 32 2 8
2 4 23 2 38 8 8
2 2
a bb
a ba
a bb
a ba
x
a bE x mean x x a b dx x x a b dxb a
a ab bE x x x a b dx x x a b dxb a
a ab b b aa ab bE x E x Variance
b aσ
+
+
+
+
⎡ ⎤− +⎢ ⎥= = − + − − + =⎢ ⎥− ⎢ ⎥⎣ ⎦
⎡ ⎤− + +⎢ ⎥⎡ ⎤ = − + − − + =⎣ ⎦ ⎢ ⎥− ⎢ ⎥⎣ ⎦
+ + −+ +⎡ ⎤ − = = − =⎣ ⎦
−=
∫ ∫
∫ ∫
![Page 22: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/22.jpg)
Equivalent probability
Uncertainties are expressed in terms of equivalent probability
The divisor “normalises components to the same probability”
![Page 23: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/23.jpg)
Central Limit Theorem
The combined uncertainty distribution is approximately Normal
Except when one uncertainty contribution DOMINATES all others:
the combined distribution is essentially that of this dominant contribution
this single dominant uncertainty could be non- normal
![Page 24: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/24.jpg)
How many values within one standard deviation? (Normal )
For normal probability distribution, the probability that a measurement will fall within one standard deviation is 67%
ss
mean
0.67
![Page 25: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/25.jpg)
1σ, 68% confidence, k=1
2σ,
95% confidence, k=2
Sample standard deviation s is an estimate of population standard deviation σ
Normal distribution
k is “coverage factor”
![Page 26: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/26.jpg)
How many values within k standard deviations? (Normal)
k 0.25 0.5 0.75 1.0 1.25 1.5 1.75 2.0 2.5 3.0 3.5 4.0
Prob (%)
20 38 55 67 79 87 92 95.2 98.8 99.7 99.95 99.99
![Page 27: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/27.jpg)
Significance of probability distributions
• Once we have determined the shape of the output probability distribution (typically normal), we can calculate which part of the values from the set of measurement results will fall between any two chosen values
• These values determine the interval, or the margin of uncertainty. The number showing the fraction of all results falling into this interval is the confidence level.
![Page 28: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/28.jpg)
Coverage factor
• Confidence level is a number expressing degree or level of confidence in the result. Usually expressed as percentage or as a number of standard deviations.
• To express overall uncertainty at another level of confidence, re-scaling by a coverage factor k is needed.
![Page 29: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/29.jpg)
Coverage factor
Distribution divisor Confidence level
Normal 1.0 67.7%
Normal 2.0 95.5%
Normal 3.0 99.7%
Rectangular √3 100%
Triangular √6 100%
![Page 30: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/30.jpg)
Uncertainty of a computed quantity
Suppose quantity A is to be computed from two measured quantities, N and t by means of a theoretical formula.
We know uncertainty in N and uncertainty in t. What is the uncertainty in A?
![Page 31: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/31.jpg)
Uncertainty Propagation
• How to calculate the uncertainty of a computed quantity
• differential calculus is needed• combined uncertainty: are the contributing
uncertainties independent or non-independent?• General way of combining uncertainties involves
sensitivity coefficients
![Page 32: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/32.jpg)
Law of Propagation of Uncertainty (no correlations : independent)
( ) ( )
( ) ( )
( )
( ) ( ) ( ) ( )
( ) ( )
( ) ( )
2
2
2 2
2
each component may be expressed as
c i
c i
i
i ireli i i i
i
c i
c ireli
i
absolute u y u y
u y u yrelative
y y
u y
u y u xu y c u x or c
y x
absolute u y c u y
u y u xrelative c
y x
=
⎡ ⎤= ⎢ ⎥
⎣ ⎦
= =
=
⎡ ⎤= ⎢ ⎥
⎣ ⎦
∑
∑
∑
∑
![Page 33: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/33.jpg)
Law of Propagation of Uncertainty : y=f(…xi …)
( ) ( )
( ) ( ) ( )
2 2 2
2 2
2 2
:
cov ,
i i i ii
i i i j i j
c
c i i i j i j
Deviations from mean values
fy y x c xx
y c x c c x x
E y u y
u y c u x c c x x
δ δ δ δ
δ δ δ δ
δ
⎛ ⎞∂= = =⎜ ⎟∂⎝ ⎠= +
=
= +
∑ ∑ ∑
∑ ∑∑
∑ ∑∑
![Page 34: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/34.jpg)
(Summation in quadrature)
....23
22
21 etcuuuuc +++=
This rule applies where the result is found using addition or subtraction
Other versions of this rule apply where the model has multiplication or division or more complicated functions
All uncertainties must be in same units and same level of confidence
Combining standard uncertainties
Combined standard uncertainty
![Page 35: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/35.jpg)
Combined Uncertainty
Once quantities x1 ., x2 , …, xN are measured andcorresponding u1 , u2 , …, uN determinedThe standard uncertainty of the output quantity y is:
Where
are sensitivity coefficients.
( ) ( ) ( )....)( 222
2221
221 NN xucxucxucyu +++=
11 x
yc∂∂
=2
2 xyc
∂∂
=N
N xyc
∂∂
=
![Page 36: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/36.jpg)
Combining Uncertainties: Absolute or Relative?
( ) ( ) ( )
( )( )
( )( )
( )
2 2 2 2
2 2 22 2 2
2 2
2
2 ; 1
2
22 2
a b
y a b
y yc ca b
absolute u y u a u b
u y u a u ba brelativey a ba b a b
= +
∂ ∂= = = =∂ ∂
= +
⎡ ⎤ ⎡ ⎤ ⎡ ⎤= +⎢ ⎥ ⎢ ⎥ ⎢ ⎥
+ +⎣ ⎦ ⎣ ⎦ ⎣ ⎦
![Page 37: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/37.jpg)
Combining Uncertainties: Absolute or Relative?
( ) ( ) ( )
( )( )
( )( )
( )
( ) ( )
2 2 2 2 2
2 2 22 2 2 2
2 2
2 2
2
2 ; 2
4 4
4 42 2
a b
y ab
y yc b c aa b
absolute u y b u a a u b
u y u a u bb a a brelativey a bab ab
u a u ba b
=
∂ ∂= = = =∂ ∂
= +
⎡ ⎤ ⎡ ⎤ ⎡ ⎤= +⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦ ⎣ ⎦
⎡ ⎤ ⎡ ⎤= +⎢ ⎥ ⎢ ⎥⎣ ⎦ ⎣ ⎦
![Page 38: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/38.jpg)
General rule for combining independent uncertainties
V is the quantity to be calculated from measurements of x and y. Uncertainties in x and y are ux and uy respectively.
The combined uncertainty uV is:
( ) ( )22
22
yxV uyVu
xVu ⎟⎟
⎠
⎞⎜⎜⎝
⎛∂∂
+⎟⎠⎞
⎜⎝⎛∂∂
=
![Page 39: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/39.jpg)
Combining independent Uncertainties. Example
Calculate kinetic energy of an object, whose mass is measured as m=0.230±0.001kg and velocity is v=0.89 ±0.01m/s.
( ) ( )22
22
vmE uvEu
mEu ⎟
⎠⎞
⎜⎝⎛∂∂
+⎟⎠⎞
⎜⎝⎛∂∂
=
2
21 mvE = ( ) 091.0
289.0230.0
21 2
2 =⋅
=mv
![Page 40: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/40.jpg)
⎥⎦
⎤⎢⎣
⎡=
⋅ Js
mkg2
2
Combining independent Uncertainties. Example
( ) ( ) 002.001.004.0001.063.0 22 =⋅+⋅=Eu
63.04
42
==⎟⎠⎞
⎜⎝⎛∂∂ vmE
mvvmvv
mmvvv
E==
∂∂
=⎟⎠⎞
⎜⎝⎛
∂∂
=∂∂ 2
21
21
21 22
E=0.091 ±0.002
222
21
21
21 vm
mvmv
mmE
=∂∂
=⎟⎠⎞
⎜⎝⎛
∂∂
=∂∂
04.02
=⎟⎠⎞
⎜⎝⎛∂∂
vE
![Page 41: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/41.jpg)
General rule for combining correlated uncertainties
Suppose x and y are completely correlated(only possible when experimental uncertainty is
negligible and the quantities are in error due to some common cause eg. Incorrectly calibrated instrument)
yxV uyVu
xVu
∂∂
+∂∂
=
![Page 42: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/42.jpg)
Graphical representation of combining uncertainties
Correlated uncertainties
uV
cx
ux
cy
uy
uV
cx
ux cy
uy
Independent uncertainties
![Page 43: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/43.jpg)
A more realistic example
![Page 44: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/44.jpg)
Practical tips for calculating sensitivity coefficients
Calculation of c1 , c2 , …. ,cN by partial differentiation is often lengthy process.
Alternatively:• Change one of the input variables by a known
amount • Keep all other inputs constant• Note the changes in the output quantity
![Page 45: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/45.jpg)
Tips for calculating sensitivity coefficients
hxyhxy
xyc )()( −+≈
∂∂
=
The amount by which you changed the input variable
Initial outputOutput after change
![Page 46: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/46.jpg)
Spreadsheet model ( “formulation” )Symbol
Source of uncertainty
[uncert] Value ±
Probability distribution
Divisor ci Standard uncertaintyu ± units
νi or νeff
One row for each contributing uncertainty
uc Combined standard uncertainty
Normal
U Expanded uncertainty
Normal (k=2)
![Page 47: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/47.jpg)
Spreadsheet model
The columns are:
Symbol or reference
Source of uncertainty –
brief text description of each uncertainty
Value (±)
– estimate of the uncertainty – from what info you have, e.g. “worst case limits” or “standard deviation”. Show units, e.g. °C.
Probability distribution - rectangular, normal, (or rarely others)
Symbol
Source of uncertainty
[uncert] Value ±
Probability distribution
Divisor ci Standard uncertaintyu ± units
νi or νeff
One row for each contributing uncertainty
uc Combined standard uncertainty
Normal
U Expanded uncertainty
Normal (k=2)
![Page 48: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/48.jpg)
Spreadsheet model
u – standard uncertainty a “standardised measure of uncertainty” calculated from previous columns: u = value ÷ divisor ×
ci .
νi – effective number of degrees of freedom – to do with reliability of the uncertainty estimate (the uncertainty in the uncertainty!) - sometimes ignored! Sometimes infinite! (∞)
Symbol
Source of uncertainty
[uncert] Value ±
Probability distribution
Divisor ci Standard uncertainty u ± units
νi or νeff
One row for each contributing uncertainty
uc Combined standard uncertainty
Normal
U Expanded uncertainty
Normal (k=2)
Divisor -
factor to normalise a value to a standard uncertainty (depends on probability distribution).
ci - sensitivity coefficient to convert to consistent units
![Page 49: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/49.jpg)
Spreadsheet modelThe rows are:
Title row
One row for each uncertainty
One row for combined standard uncertainty, uc , by summing “in quadrature” and taking square root, i.e., uc = ...222 +++ cbaFinal row showing expanded uncertainty, U = k ×
uc .
Normally coverage factor k = 2 (level of confidence of 95 % percent, if many degrees of freedom).
Expanded uncertainty is what finally should be reported.
Symbol
Source of uncertainty
[uncert] Value ±
Probability distribution
Divisor ci Standard uncertainty u ± units
νi or νeff
One row for each contributing uncertainty
uc Combined standard uncertainty
Normal
U Expanded uncertainty
Normal (k=2)
![Page 50: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/50.jpg)
Sources of uncertainty
Type A uncertainties (evaluated by the statistical analysis of a series of observations)
Term Source of uncertainty ±
u(xi
) ±
% Distribution Divisor ci ui νi
SD Repeatability 0.00554 pA 0.0727 Normal 1 1 0.0727 N-1
Type B uncertainties (evaluated by means other than the statistical analysis of a series of observations)
Term Source of uncertainty ±
u(xi
) ±
% Distribution Divisor ci ui νi
CF0Calibration factor of the principal
radionuclide 0.1042 1.00 Normal 1.00 1 1.00 ∞
CFiCalibration factor(s) of the
contaminant radionuclide N/A N/A Normal 1 N/A N/A ∞
VC Volume correction factor 0.003 0.003 Normal 1 1 0.003 ∞
XiFractional content of contaminant
radionuclide N/A N/A Normal 1 N/A N/A ∞
CfCapacitance of the feedback
capacitor 0.0125 0.0125 Normal 1 1 0.0125 ∞
DF Decay correction 1.3×10-5 0.0017 Normal 1 2.038 0.003 ∞
dV/dtVoltage reading across the
capacitor over the elapsed time
500 fA (ppm) 0.05 Normal 1 1 0.05∞
fR Reproducibility 1 0.1 Normal 1 1 0.1 ∞
BG Background 0.0015 0.006 Normal 1 2×10-4 1×10-6 ∞
W Weight of the sample 0.006 0.006 Rectangular 2√3 1 0.02 ∞
![Page 51: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/51.jpg)
Write down the measurement result and the uncertainty, (and record somewhere how you got both of these).
Express the uncertainty as:–uncertainty interval–coverage factor–level of confidence
e.g. The measured activity was 540 MBq with an expanded uncertainty of ±
2 MBq at a coverage factor k
= 2
giving a level of confidence of approximately 95 %.
How to express the answer
![Page 52: A Brief Workshop on Measurement Uncertaintyresource.npl.co.uk/docs/science_technology/ionising...A Brief Workshop on Measurement Uncertainty John Keightley National Physical Laboratory](https://reader031.vdocuments.site/reader031/viewer/2022041515/5e2aac4ff70acd66982821f5/html5/thumbnails/52.jpg)
Why use the GUM approach
• Where result “R2” depends on some result “R1” from (say) another lab, the propagation of errors law readily allows the uncertainty in R1 to be incorporated into the uncertainty of R2
• The combined standard uncertainty can be used to calculate realistic levels of confidence, due to the Central Limit Theorem.