SGS WOOL TESTING SERVICES | INFO BULLETIN
VOL 5.7
2014
ACCURACY, BIAS, PRECISION AND CONFIDENCE LIMITS
For wool testing, the term that is commonly used for these boundaries is the 95%
confidence limits (95% CL). These are the limits above and below the average result
The terms: ‘accuracy’, ‘bias’,
within which repeat measurements are expected to fall 95% of the time. In general
‘precision’ and
‘confidence
limits’ AND
are CONFIDENCE LIMITS
ACCURACY,
BIAS,
PRECISION
VOLUME 5.7
terms the 95% confidence limits are equal to twice the standard deviation of repeat
often confused. They each describe
measurements.
measurement errors, but they have
distinct meanings.
INTRODUCTION
Introduction
ACCURACY
AND BIAS
The terms: 'accuracy',
'bias', 'precision' and
'confidence limits' are often confused. They
If we were to make numerous
each describe measurement errors, but they
measurements
of the same property
have distinct meanings.
of a material, and calculate the average
Accuracy and bias
of the results, and then compare this
If we were to make numerous measurements of
average with the “true” value, we would
the same property of a material, and calculate
have
an estimate
of the
ofthis
the
the average
of the results,
and accuracy
then compare
average with the "true" value, we would have an
measurement.
estimate of the accuracy of the measurement.
The difference between our average and
The difference between our average and the
the
is sometimes
"true"“true”
value isvalue
sometimes
called the biascalled
of the
the
bias of the
isgives
a
measurement.
It ismeasurement.
a systematic effect It
and
an indication ofeffect
the extent
which the
systematic
andtogives
an indication
measurement system is out of calibration.
of the extent to which the measurement
Precision
system
is out of calibration.
Irrespective of the accuracy of a measurement
system, there is a low probability that two results
PRECISION
from that system will be completely identical,
even on the same
This variability
Irrespective
of sample.
the accuracy
of a can
be due to a variety of causes, but for any
measurement
system,
there
is a low
measurement system
in statistical
control,
the
inherent dispersion
of theresults
results isfrom
itself that
probability
that two
reproducible
Precision
is
system
willand
bemeasurable.
completely
identical,
usually measured with respect to the standard
even
onofthe
same
sample. This
variability
deviation
repeat
measurements.
In wool
testing,
thealargest
component
of
can
beusually
due to
variety
of causes,
variance
is in sampling.
but
for any
measurement system in
Confidence
limitsthe inherent dispersion
statistical
control,
of
results is system
itself reproducible
Forthe
a measurement
under statistical
control,
with a known precision,
weiscan
predict
and
measurable.
Precision
usually
the likelihood with
that repeat
measurements
will fall
measured
respect
to the standard
outside a specific set of boundaries. For wool
deviation
of repeat
measurements.
testing, the term
that is commonly
used for these
boundaries
is the 95%
confidence
(95%
In
wool testing,
usually
the limits
largest
CL). These areofthevariance
limits above
below the
component
is and
in sampling.
average result within which repeat
measurements are expected to fall 95% of the
time. In general terms
the 95% confidence limits
CONFIDENCE
LIMITS
are equal to twice the standard deviation of
repeat
measurements. system under
For
a measurement
statistical
with
a known
Maximumcontrol,
probable
differences
precision, we can predict the likelihood
When two measurements are made on the same
that
repeat
fall
lot and
are thenmeasurements
compared, there iswill
uncertainty
associatedawith
each one,
difference
outside
specific
setand
of the
boundaries.
between them is therefore affected by the
precision of both measurements. IWTO defines
the maximum probable difference (MPD) as the
difference within which 2 measurements
characterising the same lot are expected to fall
95% of the time. In practice, the MPD is equal to
1.414 times the 95% CL for the measurement.
WWW.SGS.COM
PRECISE BUT NOT ACCURATE
ACCURATE BUT NOT PRECISE
ACCURATE AND PRECISE
Determining accuracy in wool testing
Wool is a very inhomogeneous material. It is almost impossible to know the "true" value of any parameter
relating to a lot of wool. Test methods and IWTO Regulations assume that wool measurements are always
accurate, and indeed all laboratories strive for this situation. However, the reality is that a wool laboratory can
only assess its accuracy by reference to other laboratories, and must use the assumption that the average of a
number of such independent comparisons will be the best estimate of the "true" result. Hence professional
laboratories participate in routine interlaboratory trials in addition to monitoring their results against 'standard'
samples.
SGS WOOL TESTING SERVICES | INFO BULLETIN
MAXIMUM PROBABLE DIFFERENCES
When two measurements are made on
the same lot and are then compared,
there is uncertainty associated with each
one, and the difference between them
is therefore affected by the precision
of both measurements. IWTO defines
the maximum probable difference
(MPD) as the difference within which 2
measurements characterising the same
lot are expected to fall 95% of the time.
In practice, the MPD is equal to 1.414
times the 95% CL for the measurement.
DETERMINING ACCURACY IN WOOL
TESTING
Wool is a very inhomogeneous material.
It is almost impossible to know the
“true” value of any parameter relating
to a lot of wool. Test methods and
IWTO Regulations assume that wool
measurements are always accurate,
and indeed all laboratories strive for
this situation. However, the reality
is that a wool laboratory can only
assess its accuracy by reference to
other laboratories, and must use the
assumption that the average of a number
of such independent comparisons will
be the best estimate of the “true”
result. Hence professional laboratories
participate in routine interlaboratory trials
in addition to monitoring their results
against ‘standard’ samples.
FOR ENQUIRIES
Email us at [email protected]
Or contact us at:
48 Kemp Street, Kilbirnie
PO Box 15062
Wellington, New Zealand
Tel: + 64 4 387 8565
Fax: + 64 4 387 8651
www.sgs.com/agriculture
VOL 5.7
2014
PRECISION DETERMINATION IN WOOL
TESTING
The determination of precision for
most wool test methods is carried out
in accordance with standard statistical
practices. When a test method is
drafted, at least one (but usually more)
interlaboratory trials are carried out, in
which a number of participants use the
standard method on blind replicates of
a range of samples. Statistical analyses
of these trials allows estimation of the
variability both between and within
laboratories, which are then used to
calculate the 95% CL and MPD values.
These values are published in most test
methods. For wool they often vary with
the level of the property being measured.
Laboratories are able to monitor their
performance against these precision
values by carrying out replicate testing
both within their laboratory, and in
interlaboratory trials.
© 2014 SGS. All rights reserved. The information
contained herein is provided “as is” and SGS does not
warrant that it will be error-free or will meet any particular
criteria of performance or quality. Do not quote or refer
any information herein without SGS’ prior written consent.
Any unautho-rized alteration, forgery or falsification of
the content or appearance of this document is unlawful
and offenders may be prosecuted to the fullest extent of
the law.