ISO/IEC 17025:2017 Decision Rules and their Application to Meeting the Requirements

Summary

このウェビナーでは、Perry Johnson Laboratory Accreditationのマイケル・クレイマーが、ISO 17025:2017の要件を満たすための決定規則とその応用について説明します。彼は、測定不確かさを考慮に入れる方法と、顧客との合意の重要性に焦点を当て、さまざまな決定規則の例とそれらがどのように適用されるかを紹介します。また、リスク管理の新たな要求事項と、ラボが柔軟で顧客のニーズを満たすための手順についても触れています。質問応答セッションを通じて、決定規則の概念に関する参加者からの疑問に答え、次回のウェビナーについても紹介しています。

Takeaways

• 🔑 ISO 17025:2017では、測定の不確かさを考慮して適合性の判定を行う際の「決定規則」を定義し、顧客と合意する必要がある。
• 🤝 決定規則は、契約レビュー段階で顧客と合意し、文書化する必要がある。
• 🎯 決定規則には、単純な合否判定だけでなく、条件付き合格や不合格などの複数の判定レベルを設定できる。
• ⚖️ 決定規則には、偽陽性(不合格品を合格と判定)や偽陰性(合格品を不合格と判定)のリスクレベルを考慮する必要がある。
• 📚 一部の試験方法や規格では、決定規則がすでに規定されている場合があり、その場合は別途の決定規則は不要。
• 🔍 測定の不確かさを無視する「単純受入れ」の決定規則を適用する場合、その旨を文書化し、偽陽性や偽陰性のリスクを最大50%と見なす必要がある。
• 📝 報告書には、適用した決定規則と関連するリスクレベルを文書化する必要がある。
• 🔄 内部顧客に対しては、結果の報告を簡素化できるが、必要に応じて不確かさの影響を追跡できるよう記録を残す。
• 🧩 顧客が適合性の判定を必要としない場合は、決定規則の文書化は不要。
• ✨ 決定規則の適用は柔軟性が求められ、顧客の要求事項を満たすために各検査機関がケースバイケースで対応する必要がある。

Q & A

決定規則とは何ですか?

決定規則とは、測定の不確かさを考慮して、規格への適合または不適合の判定をする際の基準を定めるルールのことです。ISO 17025:2017規格で新たに導入されました。

なぜ決定規則が重要なのですか?

測定値だけでは規格への適合性を適切に判断できない場合があるため、測定の不確かさを考慮する必要があります。決定規則はこの不確かさを考慮した上で適合性を判定するルールを定めるものです。

決定規則はどのように適用されますか?

契約レビューの段階で顧客と決定規則について合意し、そのルールに従って結果を報告する必要があります。単純な合格/不合格判定だけでなく、条件付き合格/不合格などの複数の判定区分を設けることもできます。

決定規則は誰が定めるのですか?

試験所または校正機関が決定規則を定め、顧客と合意する必要があります。試験方法や規格によっては、決定規則が既に定められている場合もあります。

測定の不確かさを考慮しない場合はどうなりますか?

測定の不確かさを考慮しない単純な合格/不合格判定では、不適合品を合格と誤って判定する(受入れる)リスクや、適合品を不合格と誤って判定する(拒否する)リスクが高くなる可能性があります。最大で50%のリスクがあると言われています。

決定規則は校正証明書にどのように表記されますか?

決定規則の詳細は校正証明書に記載する必要があります。判定基準となる許容範囲、不確かさの扱い方、合格/不合格の判定区分、各判定区分のリスクなどを明記します。

決定規則は内部顧客にも適用されますか?

内部顧客の場合、決定規則の適用は必須ではありませんが、後から測定の不確かさを評価できるようにしておく必要があります。外部顧客には決定規則を適用し、契約時に合意を得る必要があります。

特定の試験方法に決定規則が組み込まれている場合はどうすればよいですか?

試験方法に決定規則が組み込まれている場合は、その規定に従えばよく、別途決定規則を定める必要はありません。ただし、顧客との合意は必要です。

決定規則を適用しない場合、どのようなリスクがありますか?

決定規則を適用しないと、不適合品を誤って合格と判定したり、適合品を誤って不合格と判定したりするリスクが高くなります。このようなリスクを顧客に伝え、合意を得る必要があります。

決定規則の選択や適用に関するガイダンスはありますか?

ILACのガイダンス文書G8には、決定規則の選択や適用に関するガイダンスが記載されています。さまざまな事例や計算例が示されているので、参考になります。

Outlines

00:00

😀 開会の挨拶と導入

スピーカーのMichael Kramerによる開会の挨拶とwebinarの目的の説明。本webinarはPerry Johnson Laboratory Accreditationから行われており、参加者にはwebinarの録画とスライドが後から提供される予定。本webinarのトピックはISO 17025:2017における決定規則とその適用であり、主に校正プログラムについての内容になる。

05:01

🧐 決定規則の要求事項

ISO 17025:2017では、顧客が規格や仕様への適合性の判定を要求した場合、その決定規則を明確に定義し、顧客と合意する必要がある。決定規則は契約レビューの際に協議され、同意を得る必要がある。決定規則は、測定の不確かさを考慮に入れる方法を説明するものである。顧客の要求に柔軟に対応できるよう、手順を定めることも規定されている。

10:01

📚 ILACガイダンス文書について

決定規則に関するILACのガイダンス文書G8が最近発行された。この文書は、ISO 17025:2017の規格や仕様への適合性の判定に際して、どのように決定規則を適用するかについてのガイダンスを提供している。決定規則の定義と、ガードバンド、簡易受入れ、テスト不確かさ比(TUR)などの関連用語についても説明されている。

15:03

🔍 規格や試験方法に決定規則が含まれている場合

一部の規格や試験方法には、既に決定規則が組み込まれている場合がある。例えばASTM E84、ASTM A29などでは、手法の中で決定規則が定義されており、さらにガードバンドを追加する必要がない。ISO 14253などの規格でもガードバンドが組み込まれている。このような場合は、規格や試験方法に従えば良い。

20:07

📐 決定規則の要件

ISO 17025:2017では、適合性の判定において測定の不確かさを考慮に入れることを求めている。通常は95%信頼の下限値(k=2)が用いられるが、場合によっては99%信頼の下限値(k=3)を要求されることもある。測定値と許容限界の関係によって、単純受入れ、不適合、条件付き適合などの判定結果が生じる。

25:08

💯 単純受入れの決定規則

顧客が単純受入れの決定規則を希望し、合意した場合、測定値が許容限界内にあれば適合、外にあれば不適合と判定する。この場合、測定の不確かさは考慮されず、偽認証や偽棄却のリスクが最大50%となることを顧客に伝える必要がある。

30:09

🎯 ガードバンドを用いた決定規則

顧客がガードバンドを用いた決定規則を希望し、合意した場合、許容限界の内側にガードバンドが設けられ、適合、条件付き適合、条件付き不適合、不適合の4種類の判定が行われる。この場合、偽認証と偽棄却のリスクは2.5%以下に抑えられるが、許容限界付近では最大50%のリスクとなる。

35:11

✅ その他の決定規則の表記例

顧客と合意の上で、測定不確かさを考慮した上で、偽認証/棄却のリスクが5%を超えない場合のみ「適合」と報告する決定規則を採用することも可能である。単純で明確な表記方法として、「測定値±測定不確かさ」が仕様内に完全に収まる場合のみ適合と判定するという決定規則もある。

40:11

⁉️ 決定規則が必要ない場合

適合性の判定が必要ない場合は、決定規則を報告する必要はない。しかし、顧客から適合性の判定が求められた際は、契約レビューの際に決定規則について合意を得る必要がある。また、一部の試験では決定規則が不要な場合もあるが、その場合は顧客との合意が重要である。

45:24

🔚 まとめと質疑応答

プレゼンテーションのまとめと、参加者からの質問に対する回答。決定規則を適用しない場合は、顧客との合意が重要であること、決定規則は契約レビュー時に合意する必要があること、内部顧客には簡易報告でも構わないことなどが強調された。特定の試験方法については回答が難しい場合があるが、ILACのガイダンス文書を参照するとよいことが説明された。

Keywords

💡決定規則

決定規則とは、校正や試験の結果について、規格への適合性を判定する際に測定の不確かさをどのように考慮するかを定めた規則のことです。この規則は、ISO 17025:2017の要求事項として新たに追加されました。規則の内容は顧客と合意して決め、校正証明書などに記載する必要があります。たとえば、単純に合格/不合格と判定するのか、条件付き合格などの判定基準を設けるのかを定める決定規則は、不確かさの取り扱いに影響し、顧客への報告にも関係します。

💡測定の不確かさ

測定の不確かさとは、測定値から真値への偏りの程度を示す指標です。決定規則では、測定値と規格限界値の関係だけでなく、この不確かさがどのように判定に影響するかを考慮する必要があります。たとえば、測定値が限界値内でも、不確かさを加味すると限界値を超える可能性がある場合、その取り扱いを決定規則で規定します。不確かさは校正証明書などに記載され、顧客が合否判定の際に参考にできます。スクリプトでは、不確かさを考慮した複数の判定区分の例が示されています。

💡合否判定

合否判定とは、校正や試験の結果について、規格への適合性を判定することです。合格/不合格の二値判定だけでなく、条件付き合格や条件付き不合格など、複数の判定区分を設けることができます。決定規則は、どのような判定区分を設け、その際の不確かさの取り扱いをどうするかを規定するものです。スクリプトでは、単純な合格/不合格の判定だけでなく、不確かさを考慮した条件付き判定の例が示されています。さらに、虚偽の合格/不合格のリスクも決定規則に含める必要があると説明されています。

💡契約レビュー

契約レビューとは、顧客との契約内容を確認し、要求事項を満たすことができるかを確認する工程です。決定規則に関しては、顧客が校正証明書などに合否判定を記載することを求める場合、この契約レビューの段階で決定規則の内容を顧客と合意し、記録しておく必要があります。スクリプトでは、契約レビューにおける決定規則の取り決めが重要な要求事項として強調されています。

💡リスク

リスクとは、ある事象が起きる確率と、その事象が及ぼす影響の組み合わせで表されます。決定規則の文脈では、虚偽の合格や不合格が問題となるリスクを考慮する必要があります。これらのリスクは決定規則に記載し、顧客と合意しておくことが求められています。たとえば、簡単な合格/不合格の判定では、虚偽の合格や不合格のリスクが50%にもなる場合があるため、そのリスクを認識しておく必要があります。スクリプトでは、リスクがこうした決定規則の重要な要素として説明されています。

💡ガードバンド

ガードバンドとは、規格の限界値の内側に設けられた判定区間のことです。測定の不確かさを考慮して、限界値よりある程度内側の値を合格、外側の値を不合格と判定するための区間です。ガードバンドを設けることで、単に値が限界値内外かだけでなく、不確かさも加味した合理的な判定ができます。スクリプトでは、ガードバンドを設けて条件付き合格/不合格の区間を作る決定規則の例が示されています。

💡IALGガイダンス文書

IALGガイダンス文書とは、国際試験所認定機関協力体(ILAC)が発行した決定規則の適用に関するガイダンス文書です。ISO 17025:2017の新しい要求事項に対応するため、決定規則の設定方法や考え方が解説されています。スクリプトでは、このIALGガイダンス文書G8を参考資料として紹介し、その内容を一部引用しながら決定規則の具体例を説明しています。このガイダンス文書は、決定規則の理解と実践に役立つ重要な資料となっています。

💡規格の引用

規格の引用とは、決定規則がその試験や校正の規格・基準に既に規定されている場合を指します。たとえば、ASTM規格などの一部の試験方法には、判定の仕方や不確かさの取り扱いが規定されていることがあります。このような場合は、その規格に従った決定規則を適用すれば良く、ISO 17025の要求事項を満たすことができます。スクリプトでは、こうした規格への引用が決定規則の設定を不要とする例外事項として説明されています。

💡内部顧客

内部顧客とは、同じ組織内の部門や関連会社などを指します。ISO 17025の要求事項は主に外部顧客向けのサービスを想定していますが、内部顧客向けのサービスについては柔軟な対応が認められています。決定規則に関しても、内部顧客の場合は必ずしも報告書に記載する必要はありません。ただし、必要に応じて不確かさを考慮した合否判定ができるよう、根拠となるデータは保持しておく必要があります。スクリプトでは、内部顧客への報告に関する質問に対してこの点が説明されています。

💡試験手順

試験手順とは、試験を実施する際に従う具体的な方法や手順のことです。一部の試験手順では、判定基準や不確かさの取り扱いが規定されている場合があります。このような場合は、その試験手順に従った決定規則を適用すれば、ISO 17025の要求事項を満たすことができます。しかし、試験の種類によっては手順に判定基準がない場合もあり、そのような場合は別途決定規則を設ける必要があります。スクリプトでは、試験手順に決定規則が含まれている例が紹介されています。

Highlights

Michael Kramer introduces the webinar on behalf of Perry Johnson Laboratory Accreditation, focusing on the decision rule and its applications to meet ISO 17025:2017 requirements.

The webinar emphasizes the importance of decision rules in the context of ISO 17025:2017, highlighting their role in accounting for measurement uncertainty in statements of conformity.

Introduction to the concept of decision rules, explaining how they have always been a requirement but were not previously defined until the 2017 standard.

Explains the necessity of defining and agreeing on decision rules during the contract review process with customers.

Discussion on whether a default decision rule can be applied and how it must be presented and agreed upon during contract review.

Highlighting the flexibility required in applying decision rules to accommodate various customer requirements and the diverse applications of tests and calibrations.

Illustration of how different applications (e.g., DNA testing, public health, educational purposes) require different levels of risk acceptance in decision rules.

Emphasis on the laboratory's responsibility to review customer requests and ensure the selection of appropriate methods and procedures.

Introduction to guidance document ILAG G8, aimed at assisting laboratories in using decision rules for declaring conformity with specifications or standards.

Explanation of guard banding and its relevance to decision rules, including the concepts of simple acceptance and the test uncertainty ratio (TUR).

Discussion on the expanded role of measurement uncertainty and decision rules in the 2017 version of ISO 17025.

Clarification on the application of decision rules in both testing and calibration, with examples from ASTM standards.

Exploration of different decision rule approaches, including binary and non-binary statements, and the implications for risk of false acceptance or rejection.

Details on how to document decision rules on calibration or test reports, emphasizing the importance of agreement with the customer.

Q&A session addressing specific questions about decision rules, measurement uncertainty, and reporting requirements.

Conclusion of the webinar with a reminder about the availability of the recorded session and slides on Perry Johnson's website.

Transcripts

hello welcome my name is Michael Kramer

I am presenting this month's webinar on

behalf of Perry Johnson laboratory

accreditation I want to thank everybody

for logging in and listening to us this

afternoon or this morning depending on

where you're happy to be located

I am the calibration inspection program

and repeat in LA and this month I will

be presenting on the decision rule and

their applications to meet the

requirements of ISO 17025

2017 they're with me I'm having a little

problem getting to my starting point

all right sorry about that folks

this webinar is being recorded all the

webinars are recorded you can go online

at the Perry Johnson's website PG laughs

comm go to the tide recorded webinars

surely when we're completed here and

that was a Friday have some colleagues

that may have started their weekend a

little are only had perhaps would

appreciate the fact they can go on our

website and listen to today's

presentation also the individual slide

will also be available on our website go

to the webinars tab and you will see the

time for just a slide presentations and

there's no need to email me and asking

for the slides they are available upon

our website along with the recording as

always that you can answer you can ask

questions during the course of this

webinar we'll entertain them at the

conclusion you should have a box there

on your screen again please keep your

questions related to today's topic

today's topic a decision rule it's going

to vary on how that's looked at

particularly among you folks in testing

just bear in mind I am a calibration

individual if your very program specific

in regards to say perhaps a specific

tester doing I may not be able to answer

your question hopefully I will answer a

lot of those during the course of this

presentation so a decision rule so what

is that should be nothing new there was

a requirement in the 2017 and

incorporations decision rule it was

there however it was never defined so

the 2017 standard is one of the

definitions that that the standard has

encompassed and you can see the

definition of a decision rule is a role

that describes how measurement

uncertainty is accounted for when making

statements of conformity with a

specified requirement when I said that

was always been a requirement we have an

Clause from the 2005 standard

underreporting so whenever you made that

statement of compliance the measurement

excuse me the uncertainty of measurement

shall be taken into account X that's

what the decision rule is as you can see

by the definition it's how it's taking

into account so you'll see it's a

magnified in 2017

so the decision rule first makes its

appearance in the 2017 after that

definition the ones in the section 4

review request tenders and contracts so

under clause 7 1 3 when the customer

requests a statement of conformity to a

specification or a standard for a test

or calibration and what we're saying

then you'll see this further clarified

in this presentation if your report is

making a statement such as pass/fail

intolerance out of Tolerance that's a

statement of conformity the decision

rule shall be clearly defined unless it

heard it in the requested specification

or standard the decision rule shall be

communicated to and a key word here is

an agreed with the customer so the key

is also when it's requested which

implies it is requested by the customer

this means that the contract review must

must take place and a clear indication

agreed on before the job is started so

instant decision rule should be

discussed and agreed to up in the nor

any a contract review process

so I proposed the question can there be

a default decision role taking or leave

and sort of approach and I have good and

we're doing more and more assessments

now of course to the 2017 standard folks

are transition tune is going to expire

in roughly a year so most everyone who

is having their assessments now when

we're coming on site is to the new

standard so I've seen this before a

decision wrong with default decision

rule prevent it during contract remote

review and of course I would say if a

customer sees that and agree and agrees

to that then you're fulfilling the

requirement during contract review that

is presented and agree to one thing I

like to specify here with the decision

rule however laboratory you'll see here

some clauses where the laboratory needs

to be flexible and if they have the

capabilities to meet the requirements of

the customers

this is specified in other parts of the

standard also I just want to have an

example here so uh what size may not fit

at all so you may have a varied array of

clients so it's about an example here of

a pipette so under these various

examples here a pipette can be used for

DNA testing in a crime lab - yeah like

that can be used for it's uh could be

used in a court of law in other words

the results of any sort of testing that

that pipette was used for public health

the pipette that can be used to test

drinking water for example or else a

pipette can be used perhaps in a

chemistry class or some other rough

estimation of delivered volumes so I

would say

based on these circumstances the left

excuse me the rebel of list of false

acceptance would not be the same and

some of your customers may have very

very different criteria for defining how

the decision rule is applied in

reporting their statement of compliance

so again the standard is also as far as

meeting customers requirements being

flexible the lavatory shall have a

procedure for reviewing request tenders

in contracts does wherever the decision

rule first appears and under the the

procedure shall ensure that the

appropriate methods and procedures are

selected and are capable of meeting the

customers requirements also the lavatory

shall cooperate with customers

representatives even clarifying the

customer's request and in monitoring the

laboratories performance in relation to

the work performed under structural

requirement so the habit reactivity

shall be carried out in such a way as to

meet the requirements of this documents

in the laboratories customers regulatory

authorities and organizations providing

recognition laboratory manner in which a

community shall ensure the communication

take place regarding the effectiveness

the management system and the importance

of meeting customers and other

requirements and also under reporting

the results shall include all

information agreed with the customer and

necessary for the interpretation of the

results and all information required by

the method used also some flexibility

may be needed in regards to risk which

is now a key factor and actually an enum

of section within the 17.25 standard so

custom

which operate both 1/3 1/7 go to five

and also you may have clients that are

now presented wrists in ISO 9001 2015

where they're looking at that that risk

perhaps more closely the risk of false

acceptance of making a statement of

compliance on their tests or calibration

reports so they're on 9001 six point one

that's where that there's require miss

Norris actions taken to address risk and

opportunities shall be proportional to

the potential impact on the conformity

of the products and services and of

course on the 1 702 5 Section 85 is a

new requirements concerning risk so if

you happen if you're at calibration lab

and you're doing working work perhaps

for a testing lab they don't know how as

per Section eight five the laboratory

shall consider the risk and

opportunities associated with laboratory

activities so well shortly after I began

writing this presentation a guidance

document came out so I want to bring

your attention to this I have

incorporated some of the ideas behind

this guidance document into my

presentation today this is an I'll a

guidance document G documents G stands

for guidance they're different than the

P documents which are policy documents

which pjl AR PL 1 2 P L 3 PL 4 stem from

I lack policies this is a guidance

document I was able to get hold of this

and review it just hot off the press so

this guidance document has been prepared

to assist laboratories in the use of

decision rules when declaring statement

of conformity to a specification or

standard

that's required now in 175 the revised

one 705 2017 recognizes that no single

decision rule can address all statements

of conformity across the Bombers scope

of testing and calibration and 175 is

for testing and calibration and there's

multiple types of testing the

calibration of course that folks under

our accreditation are accredited to

perform so a couple definitions I did

these are brought forth from the

guidance document it comes up but often

with decision rules the concept of a

guard banding the interval between

e-collar aslam it and a corresponding

accept lease limit where that accept

this living is the tolerance limit with

an acceptance limit which is guard

banded there's something called simple

acceptance this is a decision rule in

which the excess subject acceptance

limit is the same as the tolerance

limits so if you look at that and you

think about that that implies no guard

Manning that is called simple excuse me

simple acceptance term you see often

used a lot with uncertainty and

tolerance is a TR ratio test uncertainty

ratio and that is the ratio of the

tolerance of a measured quantity divided

by this is assuming your K equals to 95%

which those born cases they are divided

by the 95% expanding measurement

uncertainty of the measurement process

there's your formula they're left on at

the bottom of this slide for those folks

that are interested you can always

google it

the I like g8 document is free you could

get it right off the I lack website i do

go over some examples here I don't cover

the whole document so if you like to

that is a resource there that I'm

presenting here and there is the direct

link to all the guys excuse me to all

the I like guidance document and from

that page you can of course go directly

to g8 you might find some other

interesting documents there as well so

of course one seven oh two five and

includes criteria related to decision

rules in conformity with requirements

and resources and processes or lady to

we'll see in personnel we've already

talked about contract review it's

addressed right up in the front there

that's media agreed to up front and

reporting is described below again I

have the decision rule definition here

which is the clause three point seven in

the standard under personnel six to six

it requires that old laboratory shall

authorized personnel to perform and now

this results including statements of

conformity or opinions and

interpretation and son this individual

who's authorized to analyze these

statements that conformance should be

versed in the decision rule and of

course there are applications and of

course reporting so we basically saw in

the old 2005 standard when you make a

statement of conformity measurement

uncertainty had to be taken into account

so we'll see some I haven't per made

them here from the standard how that's

expanded particularly in the reporting

of results so when a statement of

conformity to us

specification or standard for a test

calibration is provided the laboratory

shall document the decision were

employed taking into account the level

of risks such as false accept and false

accept statistical assumptions

associated with the decision rule

employed and apply the decision rule 786

- requires that lavatory shall report on

a statement of conformity such as the

statement clearly identified over which

results the statement of conformity

applies which specification standards

are part thereof or met or not met and

then last but not least the decision

role unless it's inherited in the

requested specification or standard so

you can see yeah quite clearly here the

concept of taking measurement

uncertainty into account they greatly

expanded into 2017 standard one thing I

want to expand along these next two

slides if you look at clause 786 to see

this general rule applies and we have in

parentheses there unless isn't heard and

in the requested specification or

standards and those folks that are

involved in testing often this is a

Patrick all the various testing types of

tests and decision rule actually

incorporate it within the procedure only

the NC stays the laboratory shall report

on the statement and conformities such

as the statement clearly identifies it

is a senior rule so what does this mean

unless isn't hurried in the requested

specification or standard I have a few

samples here sure there's quite a few

others but for example there are testing

methods that determine how the rules are

applied one good common illustration is

ASTM e 84 Rockwell artists for the

testing and calibration decision rules

take uncertainty into account

effectively and the repeat testing in

other limits as a spread of the data

etc and the rules are defined within the

methods another example in is ASTM a 29

which is standard specification for

general requirements of steel or carbon

and alloy hot Walt where it is an an

auxilary table that is based on the

method on 30 to give some extra rooms to

make a decision and this comes right

back I like guidance document g9 stays

if your application already has

measurement decision rules governed by a

public standard guidance document just

some examples here ISO 1 4 2 5 3 8 6 5 5

6500 8 etc generally in those cases case

standard test methods are prescribed and

off its compliance limits already have

garbade's built into limits so any

further garb aiding to limit risk is not

necessary like I said this is that many

encompass some of you testing folks out

there may be already handled within the

test method calibration I believe is

pretty straightforward as far as a

decision rule

reporting how it should be agreed to and

then of course documented on the

calibration report is stated in the one

702 five standard sewing for those folks

even on the test again even if you

perhaps are not so much having to handle

each individual test with garb any

perhaps you have tables where it's

already built in you may also be on the

other end where you have equipment that

you're taking out to have calibrated

these mean traceable calibration heads

you need even if that's already built

into your your method if your your

encompassing that you may encompass

decision rules and have it to agree to a

decision rule with your calibration

provider now so a decision rule for

proving conformance or nonconformity

with a specification makes a

differential or other conformance or non

conformance shall be determined with a

high probability so typically will see a

K equals two or a 95% confidence

interval associated with uncertainty and

often documented on the calibration

report the expanded measurement

uncertainty you and a confident level of

across we ninety-five percent expansion

factor K equals two will genuinely be

considers as adequate there may be cases

that would require a higher confidence

level 99 percent for example and and by

prior employment I have some cases I do

recall where the folks who I was

providing calibrations for or very

versed on measurement measurement

uncertainty at the K equals two and

there were some cases I can recall

mostly dealing with those that are used

in a court of law such as that slide

with the DNA testing crime labs in other

words requiring that K equals 3 or a 99%

k factor so typically with the

measurement uncertainty is typically

calleb is typically reported at a 95%

confidence interval which is what that

example is doing at the body because

he's very close is taking into account

the t-distribution degrees of freedom

and it multiplies that expanding

uncertainty times this this case the T

value of 2 giving you that 95%

measurement uncertainty plus or minus

95% confidence interval so what are we

talking about here I mean this might be

the very basic for some folks that are

listening but when we're talking about

taking measurement into account here's a

basic case where we have a description

description here what we have for

different cases so you can see case ABCD

we have that triangular type is the

actual value and this is on either side

that's that gray area of uncertainty so

typically say that case 1 and what the

upper and lower limits are representing

here or the tolerance so for something

to be saying within specs it's got to be

within the upper and lower limit so that

first case second case if you look at

the values only the values are clearly

with him

if you look at case C and D if you just

look at the values only they were

clearly out of sex so when we look at

uncertainty you seek uncertainty aids

around the value more effects the

statement he complies when those values

are close to the tolerance limits so for

example case B the value is within

however the uncertainty that gray area

actually strays to it being out of both

the upper and lower limit they KC also

now use in clearly if you're basing your

decision just bare just based on that

value you would say that was out of

Tolerance Hale does not pass however

taking measurement uncertainty into

account it can actually if you're just

basing it on a value you can actually

make it a statement of false except

where measurement uncertainty would

affect that statement of compliance case

D is clearly out of tolerance value eyes

and also take your measurement

uncertainty into account

so I have some examples here from the I

like guidance document here where it

talks about a binary decision rule that

is a eith where you make a decision

based on a simple pass and fail so you

see here and this is for simple

acceptance if you recall simple

acceptance is where we are basically

we're not thinking measurement

uncertainty into account so in this case

if this is you're doing a primaries that

statement of simple acceptance rule

without Gore banding all four of those

cases you clearly state a pass fail so

if we look at the second one over for

example we must think that as passing

because the value is within

specifications however taking the

measurement uncertainty into account

they actually affect that decision again

with this if you're doing this again

this needs to be accepted to and agreed

to by your customer upfront during

contract review so this is just a simple

pass/fail AC on the measurement value

without regard to guard banding again

this is further depicting even if you

are and this is just taking that

previous example a little further if you

are guard banding and you're doing

simple acceptance you can see even with

the GAR banding you still specify just

those two decisions either a pass or a

fail

here are examples of a non-binary

statement with guard Mandy

so in this and these instances decision

rules exist when multiple terms may

express the results in this case here we

have conditions that could be a clear

pass a conditional pass a conditional

fail in a straight out fail so you can

read the documentation there what we're

referring to but basically what this is

doing is for example under the

conditional pass that's a statement you

would refer to on your report in this

instance you're not stating his pass

you're referring to it as a conditional

pass we're going to look at the waste

document is here in a little bit as far

as that this is a real statement looking

at the conditional pass we have a Garbi

in there so our value it's not it's

between the guard being where we are

beating against uncertainty and the

actual tolerance limit so that's the

case where it may pass but it may not

pass if you take measurement uncertainty

to account hits you can coined a phrase

conditional pass for those of you all

who is accredited with us we actually

address this currently we have not

revised any of our policies of Perry

Johnson right now in regards to the

revision of the standard we refer to

this we do give a little guidance on

this and pl3 that's measurement

uncertainty we refer to that condition

if you like it better as possibly

terminate or fail and determinate so in

these cases here you're not going to

clearly stay

pass or fail what measurement

uncertainty affects that statement in

compliance

examples of decision rules so here is an

example if you customer has agreed to it

up front a simplified acceptance rule

which is basically you're not going baby

so if a customer agrees that a pass/fail

decision or based on acceptance limits

chosen based on simple acceptance where

you're basically you're you're basically

just looking at the tolerance limit

there the expanding measurement

uncertainty calculated is calculated

Colonel Gumm guide that unsturdy a

measurement must be less than a third of

the tolerance based on manufacturer

specs we talked about t you are that

ratio is three to one statement of

conformity or mining binary the estimate

of the measurement is assumed of a

normal distribution is specified risk is

used for the risk calculation now in

this case if you are doing a your

customer has agreed to this simple

acceptance type role the risk of false

accept is is up to 50% for measured

results excuse me up back up a little

bit it could be 50% in the case that

accepted accept excuse me I accepted

items or outside the tolerance limit our

false reject if it's outside can be up

to 50% so that's the decision rule that

would need to be documented in this case

if you're not taking measurement

uncertainty into account so where are we

coming up with 50% that is in the

extreme case it says up through there so

if you figure if you're right on the

cholera's limit as far as a pass or fail

measurement uncertainty is a value plus

or minus a degree of uncertainty that's

a 50/50 when you look at it and you

think about it that way

so the risk of loss accept is

in this case if you're using a simple

acceptance role to be up to 50%

it's the decision role with that what

need to be documented here is a

simplified acceptance decision role with

you have to take into account the risk

of false accept or reject which we'll

see here shortly that in this case is up

to 50% here we have the armaiti example

where we have more than one excuse me

more than two decisions so this is where

we have the conditional past conditional

fail and that's saying this is an

example that that you're employing in

your lab and the customer agrees that

decision rules are based on garb a and

accept this limit where your car banning

the tolerance against those against the

entity except this limits being or

baited with the uncertainty where you as

the expanded measurement uncertainty

calculated for the gum standard of

conformity or not very non-binary the

estimate of the measure is assumed to

have a normal probability distribution

in risk is used for the risk calculation

in this case the risk of acceptance

items to be outside the tolerance limit

is less than 2.5 reject percent for a

rejected items the risk to be inside the

tolerance limit is less than 2.5 percent

when the measure result is close to the

collar it's limit the risk of false

accept and reject is up to 50% then I'll

expand on that a little bit so what that

is it's a little bit confusing when you

look at that but if you say something is

in the acceptance level

and I say it's right on that limit and

this is based on a 95% confidence

interval it's plus or minus so you're

gonna go plus 2.5% into that conditional

pass zone hence that leaves an

additional 2.5% where you could be

making a decision of false acceptance

when you are basically stating something

passes again the 50% that would be in

those cases where it falls directly on

the tolerance or the acceptance limit

there so here we have four different

criteria we have accept conditional pass

conditional bail well we also have a

rejection interval so here we're gonna

have several different decision rules

that should be expressed on your

calibration report if you are calibrator

so we would specify and we're stating

something that passes the measure value

were observed in tolerance at the points

tested the specified false accept is up

to 2.5%

what identify conditional pass the

measurement values were observed in

tolerance at the point tested however a

portion of the expanded measurement

uncertainty intervals that one of the

more measure values exceeded tolerance

when the measurement result is close to

the tolerance the specified false accept

is up to 50% and then here is just the

reversal on the other side that defies

the decision rule for a conditional fail

or something might accept barely fail

however it could fall over into the past

category would be a appropriate decision

role statement to define that and there

was those values that clearly fail

values were observed out of tolerance at

the point test at the specified false

reject is up to

five percent excuse me 2.5% okay so uh a

couple other common decision rule

statements perhaps you don't like the

concept of the conditional pass/fail you

want to report a pass or a fail you just

want to garb and let's say against the

five percent excuse me the five percent

of possibility of false acceptance or

false objective rejection so I have a

couple other decision rule statements

that are perfectly okay if of course

your customer agreed to it accounting

for the uncertainty will take that taken

to me that in a 95% confidence level the

measurement result plus or minus the

expanded uncertainty equals two shall be

totally within the specification limit

and the risk of false accept / rejection

will not be greater than five percent or

even more simpler the results cannot be

reported as being in specification if

the risk of false acceptance / reject

and the customer is greater than five

percent so we have to take the level of

risk into account with these decision

rules and that's the requirements data

right there in 76112 statement of

conformity to a specification or

standard is provided the lavatory shall

document the decision trowing coin they

can add it to account take it into

account the level of risk so the risk

word pops up here such as false accept

false reject and the statistical

assumptions associated with the decision

rule employed and apply the decision

rule

okay there may be cases your might be

thinking I have a completely different

application where this you may have

customers where the uncertainty or

decision rule would be very Manute or

may not really apply to them in their

areas of testing or calibration like

anything I just want to point it out if

the word if for whatever reason these

are requirements ignoring the seven

point eight one of the very first

requirements it talks about is

simplified reporting and that's based

that is based on what agreed with the

customers a results may be reported in

this simplified way and Eve any

information listed between the clauses

seven eight - seven eighty seven that is

not reporting to the customer shall be

readily available so that was also

specified at 2005 2005 specifies a

written agreement 2017 does not I would

suggest however if for whatever reason

you're bypassing this it needs to be

agreed to up front and I would capture

that in writing

create a record then it was agreed to

between you and your customer so

sometimes this isn't that this may apply

very coarse testing destructive testing

just there are example of a furniture

test on the the far left side are saying

like a truck scale calibration thousand

pound blocks of motorized

weight cart used on a scale with 24

graduations

okay that concludes today's presentation

we will pause here for a second like I

said I hopefully have looks like I got

some questions here I will try to answer

as many as I can

hi is it also a decision rule not to

take the measurement uncertainty into

consideration in the laboratory and the

customer negotiate this in a way um

again you know I try to emphasize here

you need to agree to it upfront you know

your customers are better than I do

you need to be flexible with it there

are parts of the standard or you need to

meet your customers need but but

absolutely I'm that the key here is you

need to come to an agreement and I'm

just rereading the question and what

you're talking about here is that simple

acceptance where you're not thinking you

didn't do like you so if that is the

case however let's recall that needs to

be documented on the calibration report

that decision rule and it needs to take

into account the level of risk which we

depict there when we went over that

slide that could be up to 50% so if you

and your customer has come to that

agreement that's fine and you also need

to look at it perhaps your reporting the

uncertainty your reporting just a PI's

plan without taking it into account your

customer may very well be able to make

that decision for themselves looking at

the value looking at particularly those

that are close but determining whether

or not measurement uncertainty would

affect that decision and whether or not

that would affect

perhaps their own testing or calibration

that they do in regards to the artifact

that the perhaps that are sending to you

all

I'm trying to read this next question or

contract lavatories required to provide

statements to conformity when requested

by the client if not can a lavatory

choose not to provide this service and

therefore have no need for a decision

rule well absolutely

if you're armed and I recently came

across this on an assessment laboratory

decided that they're going to report the

values without making a statement of

compliance well the very first thing yes

if your customer agreed to that up front

and you're reporting hey this is let's

say a test weight for example this is

the value of your 1 gram test weight one

point zero zero zero zero five grams and

your uncertainty is plea 0:02 milligrams

and you are not stating a pass or fail

there then the decision there is no

decision rule to be documented on the

report that is just when you're making

that statement of compliance that is

city rule needs to be there so if for

whatever reason you are not reporting in

decision you are not reporting a

statement it complies on the report then

there is no decision rule there

[Music]

all right now actually if a customer

asked to see only the readings that fall

outside the outer tolerance limit plus

the uncertainty garlanded as a fail on

the certificate that is okay again you

have to I'm not showing I'm capturing

the whole gist of the question that's

perfectly fine except you still have to

deter key you still have to document the

decision roll with the level and it

looks like you're talking about just on

the failure side with the level of false

acceptance so if you're actually guard

banding those failures

that's that K equals 2 or 95% confidence

interval you still need to document that

level of false except excuse me of a

false rejection in that case

if the customers are internal and don't

require pal Labs is for the purpose of

the pass or fail the product and I'm

reading this exactly this pipe and if

I'm missing part of the question there

so if you have an internal customer

internal to your organization and

basically with those reporting no you

don't have to report the decision rule

that simplified you're looking at

simplified reporting their play States

or internal customers and agreed to by

external gusted customers but we're

going to backtrack a little bit the

records you have to maintain the records

where where need be go back and

determine whether or not measurement

uncertainty may actually affect a

statement it complies but for internal

customers you have a lot of wiggle room

there as long as you've been tainting

the records where you could go back and

do the reverse traceability and see if

it actually affects the result so that

would go with anything you reporting the

results for internal customers they can

all be reported in any I simplified

manner

and that's the next question and this is

an area I'm not familiar with it all as

a testing he's referring to Eliza I know

what that is but as far as a testing

protocol I'm not familiar with negative

controls are always run with unknown

sample the kits provide a clear value

after which results are declared

positive that do you think we need to

draw a decision rule for our results and

you asked me what I think I think that

sounds like one of those areas that may

be incorporated within the test method

how that that ascorbate it against but

again I really don't know in that

particular instance

we're making those statement in

compliance on the certificate is no

statement regarding the decision rule

required now if you're not making a

statement in compliance there is no

decision rule statement to be made just

bear in mind if your customer requires a

statement of compliance which yeah you

need to determine it is specified in

section 7 1 under review request tenders

and contracts if it's determined or that

they don't need a statement of

compliance on the report say the value

and the measurement uncertainty is

sufficient you're not making a statement

of compliance

hence no decision rule is needed like I

said you know your customers better than

I do and that might be fine for perhaps

99.9 percent of them all right very good

questions there again now this

presentation will be available on the

I'm still getting questions nope before

I sign off just want to we already have

next what next month's webinar which is

Friday October 25th one o'clock and

we're going to look at two through the

first two sections in the 2017 standards

also expand in a one-for-one

impartiality for to confidentiality

again this presentation is available

will be available shortly on our website

the recorded version the slides will be

there as well and again that I lack

guidance document g8 that's new that was

just released there's a lot of good

information on there so if you further

reading and there's other examples

stated in there that could all I think

is an excellent resource for

particularly you calibrators to take a

look at so again real good questions I

appreciate everybody logging in I

appreciate your interest

I look forward to doing this again next

next month again enjoy the rest of your

day