Proftest - SYKE - Guide For Participants - 2019
Proftest - SYKE - Guide For Participants - 2019
Proftest - SYKE - Guide For Participants - 2019
3 Page 1 / 23
LABORATORY FUNCTION
This guide for participants is based on the Proftest SYKE’s instructions PT2 Guide for
laboratories version 1.3 (07.01.2019). This guide has been updated based on the current
operation procedures, also technical issues are updated and clarified, e.g. Proftest SYKE's
responsibility for the correctness of the information provided by participants, the
description of the standard deviation for proficiency assessment, the fees for reporting
multiple data and for the delayed cancellation of the registration, the description for
handling of personal data, and participant registration via ProftestWEB. Further, the
formulas for D% and En scores calculation are added.
PT2, version 1.3 Page 2 / 23
LABORATORY FUNCTION
Content
1 INTRODUCTION 3
2 PROFICIENCY TESTS AS PART OF THE LABORATORY MANAGEMENT SYSTEM 3
3 SYKE AS A PROFICIENCY TEST PROVIDER 3
3.1 The SYKE Laboratory centre 3
3.2 Advisory group 3
3.3 Proficiency tests organized by Proftest SYKE 4
3.4 Confidentiality and handling of personal information 4
3.5 Participant feedback 4
3.6 Subcontracting and cooperation 5
4 ORGANISING OF PROFICIENCY TESTS AT PROFTEST SYKE 5
4.1 Planning and marketing 5
4.2 Preparing and testing the samples 6
4.3 Delivery of samples 6
4.4 Processing of results 6
4.5 Performance evaluation 6
4.6 Reporting of results 7
4.7 Costs and invoicing 7
4.8 Client support and troubleshooting 8
5 PARTICIPATION IN PROFTEST SYKE TESTS 9
5.1 Contact person 9
5.2 Registration 9
5.3 Cancellation of registration 9
5.4 Receipt of samples 10
5.5 Storing of samples 10
5.6 Analysis 10
5.7 Reporting results to Proftest SYKE 11
5.8 Preliminary results from Proftest SYKE 11
5.9 Final report 12
6 OTHER BACKGROUND DETAILS FOR PARTICIPANTS 12
7 REVISIONS AND DISTRIBUTION OF THE GUIDE 12
8 REFERENCES 13
Appendix 1. Concepts and definitions 14
Appendix 2. Statistical procedure for testing samples and processing results 16
Appendix 3. Reporting results of individual participant 23
PT2, version 1.3 Page 3 / 23
LABORATORY FUNCTION
1 INTRODUCTION
The proficiency tests (PT) and other interlaboratory comparisons (ILC) organized by
the Finnish Environment Institute (SYKE) are provided under the name of Proftest
SYKE. Most commonly organized interlaboratory comparisons by Proftest SYKE are
the proficiency tests. This guide is mainly for proficiency test participants participating
in tests for chemical analysis, but, when applicable, may also be adapted for other
interlaboratory comparisons arranged by the Proftest SYKE. ProftestWEB
(https://wwwp5.ymparisto.fi/Labtest/en) is the electronic client interface of Proftest
SYKE.
The guide aims to provide an overview of organizing the Proftest SYKE proficiency
tests, to assist the understanding of the guides issued for each separate proficiency
test.
information flow between the participants and the provider of proficiency testing. The
members of the group are informed on the Proftest SYKE website
(www.syke.fi/proftest/en).
All feedback related to the proficiency testing services is documented and exploited
when arranging future proficiency tests and improving the activities.
For each PT (and ILC), the related feedback and comments are included in the final
report of the PT (or ILC).
General feedback related to the Proftest SYKE services could be sent directly to
Director of Laboratory ([email protected]). Where necessary, disagreements
arising between the organizer and the participants are aimed to be settled through
negotiation and conciliation.
cancellation latest two weeks before the planned time of realization of the proficiency
test.
Proftest SYKE promotes the upcoming PTs and ILCs also via LinkedIn
(www.linkedin.com/in/proftestsyke/) and Eptis database (www.eptis.bam.de).
content of water), the results of homogeneity and stability tests, the uncertainty of the
assigned value, the standard deviation of results, and the long-term variation in the
former proficiency tests. The standard deviation for proficiency assessment can also
be based on the legislative requirements. Preliminary values for standard deviation
for proficiency assessment are provided in the sample letter and the values are
reviewed and finalized when processing of results.
If the standard deviation for proficiency assessment set by the provider is not
appropriate for the participant’s purpose, the participant may recalculate the z score
using the formula shown in Appendix 3.
The reliability of the assigned value is tested by comparing its uncertainty to standard
deviation for proficiency assessment (Appendix 2, part 10). The reliability of the
standard deviation for proficiency assessment is tested by comparing it to the
standard deviation of the test results (Appendix 2, part 11).
When the participant has reported their results together with the uncertainty
information, the zeta scores and their comparison to the z scores are given to the
participants as part of the preliminary results (Appendix 2, part 12).
When there are only few reported results for a measurand (n < 6), the performance
could be evaluated by the means of D% (Difference) or En (Error, normalized) scores.
D% and En scores describe the difference between the participant results and
assigned value. En score includes the expanded uncertainties of the participant result
and the assigned value.
Costs are calculated on the basis of e.g. equipment, labour, delivery, printing and
similar expenses. Usually the price is divided into basic fee (same for all participants)
and separate fees for the samples. Basic fee for participation includes the sample
delivery costs within Europe. Participants from outside Europe are kindly instructed to
contact the provider to get more information of the delivery costs.
A cost estimate is prepared for each proficiency test at the time of the preparing the
annual program and reviewed when dispatching the information letter. The estimated
costs may change, for example, if the test program is modified by the request of
participants, or due to a substantial increase in costs.
The invoices are dispatched after publishing the preliminary results. The provider
defrays the delivery costs of the damaged or missing samples while the costs of
providing and delivering additional samples must be borne by the participants. The
participant may provide several results for measurands for each proficiency
test. The participant receives separate evaluation for each additional result set
and the provider will charge an additional fee of 40 % of the basic fee for each
additional data set.
The participation fee must be paid in full when the participant has registered and
received the samples but does not deliver the test results to the organizer. Each
participant must defray its own analysis costs and the possible customs fees and
similar.
The samples are pre-tested. However, in case of sample preparation failure
observed after sample delivery, there is no charge for the participants. If possible, a
new sample will be delivered to the participant at the standard charge.
5.2 Registration
Participants register (Create order) to an open PTs/ILCs via the electronic client
interface, ProftestWEB (https://wwwp5.ymparisto.fi/labtest/en) according to the given
timetable. The interface could be found also via Proftest SYKE website
(www.syke.fi/proftest/en Current proficiency tests).
When the participant has already used ProftestWEB, username and password is
used to log in. Then the contact information is filled in automatically to the New order
form. Via the Orders page on ProftestWEB, it is also possible to register to an open
PT/ILC without login. In such case, after the order is sent, the provider gives the
participant access to the interface.
When registering, the participant orders the needed samples by selecting them on
the order form. The participant may order several samples if needed. The cost for
samples is indicated on the order form. The participant could also deliver additional
set(s) of results. The participant is advised to contact the provider in such case, the
provider then creates multiple result forms for the participant. A supplementary
charge is added for this (see Chapter 4.7).
At the time of registration, participant should deliver the invoicing information
including the VAT number of foreign participant's institute, their own order number (if
needed), client code, or invoicing address, when it is not the address of the sample
delivery address.
The registration is accepted by Proftest SYKE and the acceptance is shown as a
date stamp on the information of the current test (Tests Orders).
5.6 Analysis
The sample letter generally includes details of the concentration range of the
measurands. Samples are analysed using the standard procedures of the participant.
When necessary, the proficiency test provider may issue special instructions for
sample pretreatment and measurements.
If the participant deviates from the instructions and recommendations issued with the
sample, then this deviation and the reason for it should be informed when reporting
the results. It is particularly important to inform the provider about the deviations from
the recommended time of analysis, as these deviations could affect the evaluation of
laboratory performance. If the participant has difficulties with the measurement
deadlines, they have to contact the provider to rearrange the timetable.
The provider requests participants to report either one test result or multiple results of
parallel analysis. Parallel testing is a repeat of the whole analysis from beginning to
end, including the sample preparation stages. When parallel results are not
requested by the provider, the participant will perform the analysis with as many
parallel tests as are normally conducted for the measurement.
The test analysis is also subject to normal quality assurance procedures.
PT2, version 1.3 Page 11 / 23
LABORATORY FUNCTION
The following appendices are usually provided with the preliminary results:
results reported by participant
when results are reported as parallel results, the preliminary results have the
mean value
result tables for individual participants (see Appendix 3 for an example)
definitions of statistical parameters
PT2, version 1.3 Page 12 / 23
LABORATORY FUNCTION
8 REFERENCES
1. SFS-EN ISO/IEC 17025, 2005. General requirements for the competence of testing
and calibration laboratories. Finnish Standard Association (SFS), Helsinki.
2. EURACHEM Guide, 2011. Selection, use and interpretation of proficiency testing
(PT) schemes (http://www.eurachem.org).
3. FINAS A2/2012. The principals for the assessment of the quality assurance and the
intercomparison practices in laboratories (in Finnish, http://www.finas.fi).
4. ISO/IEC 17043, 2010. Conformity assessment — General requirements for
proficiency testing. Finnish Standard Association (SFS), Helsinki.
5. ISO 13528, 2015. Statistical methods for use in proficiency testing by interlaboratory
comparisons.
6. ISO 5725-2, 1994. Accuracy (trueness and precision) of Measurement Methods and
Results - Part 2: Basic Method for the Determination of Repeatability and
Reproducibility of a Standard Measurement Method.
7. Thompson, M., Ellison, S.L. R., Wood, R., 2006. The International Harmonized
Protocol for the Proficiency Testing of Analytical Chemistry laboratories (IUPAC
Technical report). Pure Appl. Chem. 78: 145-196. www.iupac.org.
8. Nordtest Report TR 569, Edition 5.1, 2018. Internal Quality Control – Handbook for
Chemical Analytical Laboratories. (http://www.nordtest.info).
9. Nordtest Report TR 537, Edition 4, 2017. Handbook for calculation of measurement
uncertainty in environmental laboratories (http://www.nordtest.info).
10. Näykki, T., Virtanen, A. and Leito, I., 2012. Software support for the Nordtest method
of measurement uncertainty evaluation. Accred. Qual. Assur. 17: 603-612. MUkit
website: www.syke.fi/envical.
11. Lisinger, T.P.J., Kandler, W., Krska, R., Grasserbauer, M., 1998. The influence of
different evaluation techniques on the results of interlaboratory comparisons. Accred
Qual Assur 3: 322-327.
PT2, version 1.3 Page 14 / 23
LABORATORY FUNCTION
Replicate determination
Two or more parallel determinations, where the determination is repeated from beginning
to end (including the pre-process stages).
Reproducibility
Measurement conformity where test results are obtained using different methods, different
equipment, in different laboratories, by different operators and at intervals that are long in
relation to a single test. The reproducibility deviation is usually greater than the
repeatability deviation. It is generally used in proficiency testing schemes.
Stability
Samples remain unchanged (stable) until they are analysed.
Standard deviation for proficiency assessment
Measure of dispersion used in assessing proficiency, based on the available information.
Traceability
The relation of measured results through an unbroken chain of measurements to the
appropriate national or international standards.
Trueness
The closeness of agreement between the average value obtained from a large series of
test results and an accepted reference value.
Uncertainty of measurements
A parameter associated with the result of a measurement that characterizes the dispersion
of the values that could reasonably be attributed to the measurand.
PT2, version 1.3 Page 16 / 23
LABORATORY FUNCTION
1 ANOVA test
The ANOVA test can be used when participants report several replicate results to estimate
the standard errors within and between participant results [6].
The repeatability standard error sw (within participant results) is calculated using the
participants’ replicate results and also the between participants’ results standard error sb is
calculated. Finally, the reproducibility standard error st is calculated according to the
equation:
st sw 2 sb 2
2 Outlier tests
Outlier tests are used to identify the results that differ statistically significantly from the
other results in the data set (in practice, the values outside the 95 % confidence level).
The parallel results are tested with Cochran’s test and the deviation of the participant
result (or the mean of parallel measurements) from the data set is tested with the Grubbs
or Hampel test.
Cochran’s test
Cochran’s test is designed to assess the within-laboratory deviation, i.e. to determine
excessive discrepancies between participants [6]. Participants are numbered 1, 2, ..., p
and iterated distributions s1, s2,…, sp. The test value is:
s 2 max
C p
, where
si2
i 1
Cochran’s test is performed when there are parallel results from at least three participants
in the result data.
PT2, version 1.3 Page 17 / 23
LABORATORY FUNCTION
Grubbs test
In the Grubbs test the result deviation is tested either one by one (biggest or smallest
results, Grubbs) or two by two (biggest or smallest, Grubbs2). In the test the values are
calculated for the minimum and maximum results. For the Grubbs test the test value G is
the bigger from the results of minimum value / and maximum value
/ , where is mean of reported results, is smallest result, is
maximum result and s is standard deviation of the reported results. The Grubbs2 test
compares the variance of whole data to the variance observed when two highest or lowest
results have been eliminated from data. The result is outlier if test value G is higher than
critical value in the 5 % significance level. The Grubbs test could be repeated and applied
to the data until no more outliers have been found [6]. However, after the test at least three
valid values should remain.
Hampel test
Hampel test is based on the median and the absolute value of a single value. The median
xmed (see part 6) of the results x1, x2,…, xp is calculated together with the absolute residuals
(di) of the single results from the median (di = |xmed- xi|). The median of the absolute
residuals MAD (Median Absolute Deviation) is then calculated. The result xi is an outlier if
di > 5.06 × MAD [11].
When interpreting the results of the outlier tests, the standard deviation for proficiency
assessment (spt) is taken into account. The outlier test is performed when the data
consists of at least seven results.
Robust analysis
The use of robust statistics also allows discarding of extreme results before calculating the
final robust mean (see part 8, [5]).
3 Homogeneity test
For homogeneity testing 4–15 bottles (circa 10 % of the total amount) from the prepared
sample series are used and at least one measurand is determined.
Test results are assessed by analysing the variance between groups (ANOVA), with at
least two parallel analyses performed for each sample. Finally the F-test is used to decide
whether the discrepancies between the concentrations of measurand in different bottles
are significant [5, 7].
4 Mean
The mean value of results is calculated using the formula:
1n
x xi , where
n i1
5 Standard deviation
The standard deviation is the size of result distribution around the mean and is calculated
using the formula:
n
(x i x) 2
s i 1
n 1 , where
6 Median
The median is the middle result of a series arranged in order of ascending size (when n is
odd number) or the mean of the two middle results (when n is even).
7 Normality test
The normality of the result material is tested using the Kolmogorov-Smirnov test, where
the results x1, x2, … ,xp are combined in an empirical cumulative distribution function of the
x value. The number of results xi smaller than x is calculated and normalized by dividing by
the number of results p. The derived cumulative distribution is compared to the standard
cumulative distribution function (the maximum deviation of these is computed and
compared to the test value distribution).
{ x* - φ, if xi < x* - φ
xi* = { x* + φ, if xi > x* + φ
{ xi otherwise.
The new values of x* and s* are calculated from:
∗ ∗
∑ /
p
s * 1.134 ( xi * x * ) 2 /( p 1)
i 1
To determine the final robust estimates xrob and srob the robust mean x* and the robust
standard deviation s* may be derived by an iterative calculation, i.e. by updating the values
of x* and s* several times until the process converges.
9 Stability test
The stability of samples is tested when the analysed compound has poor stability e.g.
during transport of samples (e.g. determining pH, BOD7, chlorophyll a). Stability is tested
after keeping the samples cool (4 °C) and at room temperature (20 °C) during the period of
transport. Both samples are tested and the results are processed using the difference in
results obtained by analysing samples kept at different temperatures. The difference
should be smaller than 0.3 × standard deviation [5, 7]:
D =│c20°- c4°│< 0.3 × spt, where
If the assigned value is calculated as the robust mean, then the uncertainty is calculated
using the robust standard deviation at the 95 % confidence level [5]:
2 ∙ 1,25 ∙ /√ , where
srob = the robust standard deviation and n = the number of the results.
The standard uncertainty of the assigned value (upt) is compared to the standard deviation
for the proficiency assessment (spt) with the following criterion [7]:
upt/spt 0.3
The assigned value is reliable when the criterion is fulfilled. If 0.3 < upt/spt l, where 0.3 < l
< 0.7, then the assigned value has high uncertainty. If upt/spt > l, z scores will not be
reported [5, 7].
When metrologically traceable result (eg. ID-ICP-MS) is used as assigned value, the
standard uncertainty of the measurement (GUM calculated) is used as the standard
uncertainty of the assigned value.
, where
zeta = / , where
If the measurement uncertainty reported by a participant is realistic, then the z and zeta
scores will be similar. Neither is the discrepancy large if the difference xi - xpt is small, in
which case the result for participant will be near the assigned value. Participant
performance is not evaluated on the basis of the zeta score, but the participant could use it
when estimating the measurement uncertainty.
When the number of reported results is low (n < 6) and the uncertainty is set for the
assigned value, the performance could be estimated by means of En scores (’Error,
normalized’, Appendix 4). These are used to evaluate the difference between the assigned
value and participant’s result within their claimed expanded uncertainty. En scores are
calculated:
, where
where:
Measurand The tested parameter
z score Calculated z score(satisfactory result -2 ≤ z ≤ 2)
Assigned value: Assigned value
2×spt % Standard deviation for proficiency assessment (95 % confidence
level)
Participant's result: Result of an individual participant (when parallel results are reported,
the mean value of those) 1)
Md: Median value
s: Standard deviation (absolute)
s%: Standard deviation as percent
n (stat): Number of participants in statistical processing
1)
In performance evaluation, the z score is calculated from the precise result reported by the participant. In the result
sheet of the report, the Participant’s result might slightly differ from the reported value due to the number of visible
decimals or due to rounding.
z score:
In the example above, the assigned value for Ntot in sample B2N was 452 µg/l (= xpt) and
the standard deviation for proficiency assessment spt (2×spt %, at the 95 % confidence
level) was 15 %, thus spt = 7.5 % of the assigned value.
The result of the participant 5 was 472 µg/l (= xi)
z = (xi – xpt)/spt = (472-452) / (0.075 × 452) = 0.590.