COVER ILAC-G8:1996. Guidelines on Assessment and Reporting of Compliance with Specification (based on measurements and tests in a laboratory)

Transcription

1 COVER ILAC-G8:1996 Guidelines on Assessmen and Reporing of Compliance wih Specificaion (based on measuremens and ess in a laboraory)

2 Copyrigh ILAC 1996 ILAC publicaions may no be copied for sale by any individual or body oher han ILAC member organisaions

3 ILAC-G8:1996 Guidelines on Assessmen and Reporing of Compliance wih Specificaion (based on measuremens and ess in a laboraory)

4 ILAC-G8:1996 Guidelines on Assessmen and Reporing of Compliance wih Specificaion PREAMBLE These guidelines have been prepared o assis laboraories worldwide on he mehod for saing es resuls and assessmen of compliance wih specificaion for esing laboraories. In order o saisfy he requiremens of EN and ISO/IEC Guide 25, laboraories shall provide cliens wih saemens of es resuls, heir uncerainy, and he assessmen of compliance wih specificaion when requesed o do so wihin he scope of hese guidelines. For calculaion of uncerainy, oher inernaionally agreed documens are available 4. The homogeneiy of samples is assumed. The use of validaed es mehods and calibraed equipmen is assumed. All calibraions should, whenever possible, be raceable o naional or inernaional sandards. PURPOSE The guidelines give rules for supplier and clien concerning he assessmen and reporing of compliance or non-compliance for a single uni of produc using an agreed es mehod. Legal requiremens supersede any agreemen. AUTHORSHIP These guidelines were prepared by a Working Group of ILAC Commiee 3. The Convenor was Erik L Daae of he Neherlands. The guidelines were endorsed for publicaion by ILAC Resoluion No. 17/96. Page 4

5 TABLE Guidelines OF CONTENTS on Assessmen and Reporing of Compliance wih Specificaion ILAC-G8:1996 PREAMBLE... 4 PURPOSE... 4 AUTHORSHIP METHOD OF STATING TEST RESULTS General Approach Possibiliies of Obaining a Tes Resul Special Cases ASSESSMENT OF COMPLIANCE WITH SPECIFICATION REFERENCES APPENDIX A Page 5

6 ILAC-G8:1996 Guidelines on Assessmen and Reporing of Compliance wih Specificaion 1. METHOD OF STATING TEST RESULTS 1.1 General Approach The exen of he informaion given when reporing he es resul and is uncerainy should be relaed o he requiremens of he clien, he specificaion and he inended use of he resul. The mehods used o calculae he resul and is uncerainy should be available eiher in he repor or in he records of he es including: - sufficien documenaion of he seps and calculaions in he daa analysis o enable a repea of he calculaion if necessary; - all correcions and consans used in he analysis, and heir sources; - sufficien documenaion o show how he uncerainy is calculaed When reporing he es resul and is uncerainy, he use of excessive numbers of digis should be avoided. In mos cases he uncerainy need be expressed o no more han wo significan figures (alhough a leas one more figure should be used during he sages of esimaion and combinaion of componen uncerainies in order o minimise rounding errors) Unless oherwise specified, he es resul should be repored ogeher wih he expanded uncerainy appropriae o he approximaely 95% level of confidence, in he following manner: Measured value Uncerainy of measuremen (unis) ± 0.1 (unis) When specificaion describes an inerval wih an upper and lower limi, he raio of he uncerainy of measuremen o he specified inerval shall be reasonably small (i.e. 1:3). 1.2 Possibiliies of Obaining a Tes Resul Tesing o limi values When esing is done by comparing he es resul wih a limi value raher han measuring a specific value, he esimaion of he uncerainy is also mandaory. Depending on he definiion of he specificaion limis, he es resul should be repored in he same way as >, >, < or < Tesing resuls in specific values When esing is done wih a specific value as a resul here are wo possibiliies: - only one es sample is available (single es sample) - wo or more es samples from he same uni of produc are available (duplicae, riplicae ec. es samples). The cerainy of he average es resul is dependen on he number of samples. When he number of samples increases, he average es resul will beer approximae he rue value. Therefore he number of es samples should be noed in he repor. The (average) es resul and uncerainy of measuremen should be repored as menioned under secion The calculaion mehods of he average es resul and uncerainy are ouside he purpose of hese guidelines. Page 6

7 Guidelines on Assessmen and Reporing of Compliance wih Specificaion ILAC-G8: Special Cases In excepional cases, where a paricular facor or facors can influence he resuls, bu where he magniude can no be eiher measured or reasonably assessed, he saemen will need o include reference o ha fac Any uncerainy ha resuls from he es sample, no being fully represenaive of he single uni of produc, should normally be idenified separaely in he evaluaion of uncerainy. However, here may be insufficien informaion o enable his o be done, in which case his should be saed in he repor. A possible remark could be: The es resuls in his repor relae only o he es sample as analysed and no o he single uni of produc from which he es sample was drawn. 2. ASSESSMENT OF COMPLIANCE WITH SPECIFICATION 2.1 These guidelines require ha, when a es is carried ou o a saed specificaion and he clien or he specificaion requires a saemen of compliance, he repor mus conain a saemen indicaing wheher he es resuls show compliance wih he specificaion. There are a number of possible cases where he uncerainy has a bearing on he compliance saemen and hese are examined below. 2.2 The simples case is where he specificaion clearly saes ha he es resul, exended by he uncerainy a a given level of confidence, shall no fall ouside or wihin a defined specificaion limi or limis. In hese cases (Case 1, 5, 6 and 10 of Appendix A), assessmen of (non)compliance would be sraighforward. 2.3 More ofen, he specificaion requires a compliance saemen in he cerificae or repor bu makes no reference o aking ino accoun he effec of uncerainy on he assessmen of compliance. In such cases i may be appropriae for he user o make a judgemen of compliance, based on wheher he es resul is wihin he specified limis wih no accoun aken of he uncerainy. This is ofen referred o as shared risk since he end-user akes some of he risk ha he produc may no mee he specificaion afer being esed wih an agreed measuremen mehod. In his case here is an implici assumpion ha he uncerainy of he agreed measuremen mehod is accepable and i is imporan ha i can be evaluaed when necessary. Naional regulaions can overrule he shared risk principle and can pu he uncerainy risk on one pary. 2.4 An agreemen beween he clien and he laboraory or a code of pracice or specificaion may sae ha uncerainy can be ignored when judging compliance. Similar consideraions as for shared risk (above) apply in such circumsances. 2.5 In he absence of any crieria, es specificaions, clien s requiremens, or codes of pracice, he following approach is recommended: (a) (b) if he specificaion limis are no breached by he es resul, exended by half of he expanded uncerainy inerval a a level of confidence of 95%, hen compliance wih he specificaion can be saed (Case 1 and 6 of Appendix A); where an upper specificaion limi is exceeded by he es resul even when i is exended downwards by half of he expanded uncerainy inerval, hen noncompliance wih he specificaion can be saed (Case 5 of Appendix A); Page 7

8 ILAC-G8:1996 Guidelines on Assessmen and Reporing of Compliance wih Specificaion (c) (d) if a lower specificaion limi is breached even when he es resul is exended upwards by half of he expanded uncerainy inerval, hen non-compliance wih he specificaion can be saed (Case 10 of Appendix A); if he measured single value wihou he possibiliy of esing more samples from he same uni of produc falls sufficienly close o a specificaion limi, such ha half of he expanded uncerainy inerval overlaps he limi, i is no possible o confirm compliance or non-compliance a he saed level of confidence. The es resul and expanded uncerainy should be repored ogeher wih a saemen indicaing ha neiher compliance nor non-compliance was demonsraed. A suiable saemen o cover hese siuaions (Case 2, 4, 7 and 9 of Appendix A) would be, for example: The es resul is above (below) he specificaion limi by a margin less han he measuremen uncerainy; i is herefore no possible o sae compliance/non-compliance based on 95% level of confidence. level of less han 95% is accepable, a compliance/non-compliance saemen may be possible. If he law requires a decision anyway concerning rejecion or approval, case 2 and 7 can be saed as compliance wih he specificaion limi (wih a lower han 95% confidence level). In Case 4 and 9 of Appendix A, non-compliance wih he specificaion limi can be saed (wih a lower han 95% confidence level). If wo or more samples of a single uni of produc can be esed, replicae esing is advisable. Afer esimaing he average value for all es resuls on he same samples and he new uncerainy for his average-value, he same judgemen as above described should be made. (e) If he es resul is exacly on he specificaion limi, i is no possible o sae compliance or non-compliance a he saed level of confidence. The es resul and expanded uncerainy should be repored ogeher wih a saemen indicaing ha neiher compliance nor non-compliance was demonsraed a he saed level of confidence. A suiable saemen o cover hese siuaions (Case 3 and 8 of Appendix A) would be for example: The es resul is equal o he specificaion limi; i is herefore no possible o sae eiher compliance or non-compliance a he saed level of confidence. If he law requires a saemen concerning he assessmen in he form of compliance or non-compliance regardless of he level of confidence aking ino accoun he provisions of Secion 2.3, he saemen depends on he definiion of he specificaions: if he specificaion limi is defined as < or > and he es resul is equal o he specificaion limi, hen non-compliance can be saed. if he specificaion limi is defined as < or > and he es resul is equal o he specificaion limi, hen compliance can be saed. Page 8

9 Guidelines on Assessmen and Reporing of Compliance wih Specificaion ILAC-G8: REFERENCES 1. NIS 80 (1994), Guide o he Expression of Uncerainies in Tesing, NAMAS, Teddingon, UK 2. NF E , Verificaion des olerances des produis, Declaraion de conformie, December 1993, AFNOR, Paris, France. 3. ISO/DIS 14253, par 1, Decision rules for proving conformance or non-conformance wih specificaion, 1995, Inernaional Organisaion for Sandardizaion, Geneva, Swizerland. 4. BIPM, IEC, IFCC, ISO, IUPAC, IUPAP, OIML, Guide o he Expression of Uncerainy in Measuremen. Inernaional Organisaion for Sandardizaion, Geneva, Swizerland, ISBN , Firs Ediion, ISO 3534 Par 1, Probabiliy and General Saisical Terms , Saisics - Vocabulary and symbols, Inernaional Organisaion for Sandardizaion, Geneva, Swizerland. 6. VIM, ISO (1993), Inernaional Vocabulary of Basic and General Terms in Merology, Inernaional Organisaion for Sandardizaion, Geneva, Swizerland, ISBN , Second Ediion. Page 9

10 ILAC-G8:1996 Guidelines on Assessmen and Reporing of Compliance wih Specificaion APPENDIX A Case 1 Case 2 Case 3 Case 4 Case 5 Specified upper limi Specified lower limi resul is under he upper limi, even when exended upwards by half of he uncerainy inerval. The produc herefore complies wih he specificaion. resul is below he upper limi, bu by a margin less han half of he uncerainy inerval; i is herefore no possible o sae compliance. level of less han 95% is accepable, a compliance saemen may be possible. resul is on he limi iself; i is herefore no possible o sae compliance nor non-compliance. level of less han 95% is accepable, and he specificaion limi is defined as <, a compliance saemen may be possible. When he specificaion limi is defined as <, a noncompliance saemen may be possible. resul is above he upper limi, buy by a margin less han half of he uncerainy inerval; i is herefore no possible o sae non-compliance. level of less han 95% is accepable, a noncompliance saemen may be possible. resul is beyond he upper limi, even when exended downwards by half of he uncerainy inerval. The produc herefore does no comply wih he specificaion. Case 6 Case 7 Case 8 Case 9 Case 10 resul is above he lower limi, even when exended downwards by he half of he uncerainy inerval. The produc herefore complies wih he specificaion. = measuremen resul wih agreed mehod I = uncerainy inerval of agreed mehod resul is above he lower limi, bu by a margin less han half of he uncerainy inerval; i is herefore no possible o sae compliance. level of less han 95% is accepable, a compliance saemen may be possible. resul is on he limi iself; i is herefore no possible o sae compliance nor non-compliance. level of less han 95% is accepable, and he specificaion limi is defined as >, a compliance saemen may be possible. When he specificaion limi is defined as >, a noncompliance saemen may be possible. resul is below he lower limi, bu by a margin less han half of he uncerainy inerval; i is herefore no possible o sae non-compliance. level of less han 95% is accepable, a noncompliance saemen may be possible. resul is beyond he lower limi, even when exended upwards by half of he uncerainy inerval. The produc herefore does no comply wih he specificaion. Page 10