1、Designation: D 4483 05aStandard Practice forEvaluating Precision for Test Method Standards in theRubber and Carbon Black Manufacturing Industries1This standard is issued under the fixed designation D 4483; the number immediately following the designation indicates the year oforiginal adoption or, in
2、 the case of revision, the year of last revision. A number in parentheses indicates the year of last reapproval. Asuperscript epsilon (e) indicates an editorial change since the last revision or reapproval.INTRODUCTIONThe primary precision standard for ASTM test method standards is Practice E 691; a
3、 genericstandard that presents the fundamental statistical approach and calculation algorithms for evaluatingrepeatability and reproducibility precision. However, certain parts of Practice E 691 are notcompatible with precision as evaluated in the rubber manufacturing and carbon black industries ove
4、rthe past four decades. Thus a separate standard is required for precision in these two industries. Thispractice is being issued as a major revision of Practice D 4483, which has been used for precisionevaluation by Committee D11 since 1985. The basic Practice D 4483 precision calculation algorithms
5、,the same as in Practice E 691, are unchanged. This new revised Practice D 4483, organized toaccommodate the requirements of the rubber and carbon black manufacturing industries, has threenew features that provide for a more formal and structured analysis of interlaboratory test program(ITP) data.Fi
6、rst it addresses the overriding issues with precision evaluation over the past several decadesthefrequent discovery that reproducibility for many test methods is quite poor. Experience has shown thatfrequently poor reproducibility is caused by only a few laboratories that differ from the remainder t
7、hatgive good agreement. A new procedure designated as robust analysis provides an improved methodfor detecting outliers that cause poor precision, especially poor between laboratory agreement.Second, after outlier detection the new standard provides two options; (1) outlier deletion or (2) outlierre
8、placement. When outliers are deleted the revised standard provides a way to retain the non-outlierlaboratory data. This allows for a broader database for precision calculation. The current ASTMCommittee E11 computer program for calculating precision does not allow for outlier deletion in thisway. Th
9、ird, when exercising outlier Option 2, the standard gives a procedure for calculating specialreplacement values for deleted outliers in ITPs that have only a few participating laboratories. Thereplacement values are obtained in a way that preserves the observed data distribution of thenon-outlier da
10、ta. This is important since many ITPs are in the limited number of participatinglaboratories category.1. Scope1.1 This practice covers guidelines for evaluating precisionand serves as the governing practice for interlaboratory testprograms (ITP) used to evaluate precision for test methods asused in
11、the rubber manufacturing and the carbon black indus-tries. This practice uses the basic one way analysis of variancecalculation algorithms of Practice E 691. Although bias is notevaluated in this practice, it is an essential concept in under-standing precision evaluation.1.2 This practice applies to
12、 test methods that have testresults expressed in terms of a quantitative continuous variable.Although exceptions may occur, it is in general limited to testmethods that are fully developed and in routine use in a numberof laboratories.1.3 Two precision evaluation methods are given that aredescribed
13、as robust statistical procedures that attempt toeliminate or substantially decrease the influence of outliers.The first is a General Precision procedure intended for all testmethods in the rubber manufacturing industry, and the secondis a specific variation of the general precision proceduredesignat
14、ed as Special Precision, that applies to carbon black1This practice is under the jurisdiction ofASTM Committee D11 on Rubber andis the direct responsibility of Subcommittee D11.16 on Application of StatisticalMethods.Current edition approved Dec. 1, 2005. Published January 2006. Originallyapproved i
15、n 1985. Last previous edition approved in 2005 as D 4483 05.1Copyright ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959, United States.testing. Both of these procedures use the same uniform levelexperimental design and the Mandel h and k statistics to reviewth
16、e precision database for potential outliers. However, they useslight modifications in the procedure for rejecting incompatibledata values as outliers. The Special Precision procedure isspecific as to the number of replicates per database cell ormaterial-laboratory combination.1.4 This practice is di
17、vided into the following sections:SectionScope 1Referenced Documents 2Terminology 3Significance and Use 4Precision EvaluationGeneral Precision and Spe-cial Precision5Steps in Organizing an Interlaboratory Test Program(ITP)6Overview of the General Precision Analysis Proce-dure7General Precision: Anal
18、ysis Step 1 8Preliminary Graphical Data Review 8.1Calculation of Precision for Original Database 8.2Detection of Outliers at 5 % Significance LevelUsing h and k Statistics8.3Generation of Revision 1 Database Using OutlierTreatment Option 1 or 28.4General Precision: Analysis Step 2 9Calculation of Pr
19、ecision for Revision 1 Database 9.1Detection of Outliers at 2 % Significance LevelUsing h and k Statistics9.1Generation of Revision 2 Database Using OutlierTreatment Option 1 or 29.1.2General Precision: Analysis Step 3 10Calculation of Precision Using Revision 2 Data-base10.1Special Precision Analys
20、isCarbon Black Testing 11Format for Precision Table and Clause in TestMethod Standards12Preparation of Report for Precision Analysis 13Definitions for Selected Terms Concerned with Preci-sion and TestingAnnex A1Statistical Model for Interlaboratory Testing Pro-gramsAnnex A2Calculating the h and k Co
21、nsistency Statistics forOutliersAnnex A3Spreadsheet Calculation Formulas, Table Layout,and Calculation SequenceAnnex A4Procedure for Calculating Replacement Values ofDeleted OutliersAnnex A5Example of General Precision EvaluationMooneyViscosity TestingAnnex A61.5 Six annexes are presented; these ser
22、ve as supplements tothe main body of this practice. Annex A1 and Annex A2 aregiven mainly as background information that is important for afull understanding of precision evaluation. Annex A3-AnnexA5 contain detailed instructions and procedures needed toperform the operations as called for in variou
23、s parts of thepractice. The use of these annexes in this capacity avoids longsections of involved instruction in the main body of thispractice. This allows for a better presentation and understand-ing of the central concepts involved in the evaluation ofprecision. Annex A6 is also important; it give
24、s a completeexample of precision evaluation that illustrates all of theprocedures and options likely to be encountered in anyprecision evaluation, from the simple to the most complex.1.6 This standard does not purport to address all of thesafety concerns, if any, associated with its use. It is there
25、sponsibility of the user of this standard to establish appro-priate safety and health practices and determine the applica-bility of regulatory limitations prior to use.2. Referenced Documents2.1 ASTM Standards:2D 1646 Test Methods for RubberViscosity, Stress Relax-ation, and Pre-Vulcanization Charac
26、teristics (Mooney Vis-cometer)D 6600 Practice for Evaluating Test Sensitivity for RubberTest MethodsE 691 Practice for Conducting an Interlaboratory Study toDetermine the Precision of a Test Method2.2 ISO Standard:3ISO 289 Determination of Viscosity of Natural and Syn-thetic Rubbers by the Shearing
27、Disk Viscometer3. Terminology3.1 Anumber of specialized terms or definitions are definedin a systematic sequential order, from simple terms to complexterms. This approach allows the simple terms to be used in thedefinition of the more complex terms; it generates unambigu-ous definitions. Thus the de
28、finitions do not appear in the usualalphabetical sequence.3.1.1 This terminology section contains explanatory notesfor many of the definitions as well as discussion on theconnection between some of the terms and the various ways theterms are used in testing and precision evaluation. For specialempha
29、sis, a few terms are defined in the main text of thispractice where certain precision concepts are discussed.3.1.2 Annex A1 is included as part of this practice with twoobjectives: (1) Annex A1 presents new more comprehensivedefinitions drafted with substantial tutorial content, and (2)Annex A1 pres
30、ents some ancillary definitions that may pro-mote a better understanding of precision.3.2 Testing Terms:3.2.1 balanced uniform level design, nthe plan for aninterlaboratory test program for precision, where all laborato-ries test all the materials selected for the program and eachlaboratory conducts
31、 the same number of repeated tests, on eachmaterial.3.2.2 element, nthe entity that is tested or observed, toevaluate a property or characteristic; it may be a single objectamong a group of objects (test pieces, and so forth) or anincrement or portion of a mass (or volume) of a material.3.2.2.1 Disc
32、ussionThe generic term element has a numberof synonyms: test piece, test specimen, portion, aliquot part,subsample, and laboratory sample.3.2.3 element class (or class of elements), nthe categoryor descriptive name for a group of elements that have acommon origin or have nominally identical properti
33、es.3.2.3.1 DiscussionThe term nominally identical impliesthat the elements come from a source that is as homogeneousas possible with regard to the property being measured.2For referenced ASTM standards, visit the ASTM website, www.astm.org, orcontact ASTM Customer Service at serviceastm.org. For Ann
34、ual Book of ASTMStandards volume information, refer to the standards Document Summary page onthe ASTM website.3Available from International Organization for Standardization (ISO), 1 rue deVaremb, Case postale 56, CH-1211, Geneva 20, Switzerland.D 4483 05a23.2.4 test result, nthe value of a character
35、istic obtained bycarrying out a specified test method.3.2.4.1 DiscussionThe test method should specify thatone or a number of individual measurements, determinations,or observations be made and their average or another appro-priate function (median or other) be reported as the test result.3.2.5 test
36、ing domain, nthe location and operational con-ditions under which a test is conducted; it includes a descrip-tion of the element preparation (test sample or test piece), theinstrument(s) used (calibration, adjustments, and settings), theselected test technicians, and the surrounding environment.3.2.
37、5.1 global testing domain, na domain that encom-passes two or more locations or laboratories, domestic orinternational, typically used for producer-user testing, productacceptance, and interlaboratory test programs.3.2.5.2 local testing domain, na domain comprised of onelocation or laboratory as typ
38、ically used for quality control andinternal development or evaluation programs.3.3 Material and Sampling Terms:3.3.1 independent tests, na set of measurements (or ob-servations) for a defined testing domain, where, in relation tothe measurement process, there is no influence of any selectedmeasureme
39、nt on any other measurement in the set.3.3.1.1 DiscussionThe word independent is used through-out this practice as an adjective to indicate the concept ofindependence, for samples, test pieces, and so forth, as well astests.3.3.2 lot, na specified mass or volume of material ornumber of objects; usua
40、lly generated by an identifiable pro-cess, frequently with a recognized composition or propertyrange.3.3.2.1 DiscussionA lot may be generated by a commonproduction (or other natural) process in a restricted time periodand usually consists of a finite size or number. A lot may be afractional part of
41、a population (Interpretation 2 of population,see Annex A1). A recognized property range implies that somerough approximation is available.3.3.3 material, na specific entity or element class to betested; it usually exists in bulk form (solid, powder, or liquid).3.3.3.1 DiscussionMaterial is used as a
42、 generic term todescribe the class of elements that is tested, that is, a materialmay be a rubber, a rubber compound, a carbon black, a rubberchemical, and so forth. A material may or may not behomogeneous. In product testing the term material may be usedto describe the class of elements or type of
43、rubber productssuch as O-rings, hose assemblies, motor mounts, and so forth.See also 5.1.4.1.3.3.4 sample (data), nthe number of test or observationvalues (n = 1, 2, 3, and so forth), obtained from (one or more)physical samples, by the application of a specific test (obser-vation) method.3.3.5 sampl
44、e (physical), nthe number of elements or thespecified mass of a material, selected according to a particularprocedure, used to evaluate material, lot, or population char-acteristics.3.3.5.1 DiscussionThe term sample should not be used asa synonym for material, see 3.3.3,ortarget material, see5.1.4.1
45、. Ideally several materials are tested in any ITP witheach material being different (chemically, structurally, propertywise). From each material, some number of samples (allnominally identical) may be taken for testing. See 3.3.4.3.3.6 test sample, nthat part of a (physical) sample of anytype taken
46、for chemical or other analytical testing, usually witha prescribed blending or other protocol.3.3.6.1 DiscussionA test sample is usually a mass orvolume that is some small fractional part of a bulk material.3.3.7 test specimen, nan object (appropriately shaped andprepared) taken from a sample for ph
47、ysical or mechanicaltesting.3.3.7.1 DiscussionOther terms for test specimen are: testportion, test item, and test piece (used in ISO standards).3.4 Statistical Terms Relating to Precision:3.4.1 estimated (true or reference) mean, nthe meanobtained on the basis of n independent replicate measurements
48、;the greater n the better the approximation to the true orreference mean, provided there is no systematic deviation orbias.3.4.1.1 DiscussionThe words mean and estimated meanare frequent synonyms for estimated (true or reference) mean.The value for n in typical routine testing programs is of theorde
49、r 1 to 10. When bias exists, the estimated (true orreference) mean so obtained estimates + S Bi, where =true or reference mean and S Bi = algebraic sum of all biasdeviation terms. Therefore, if bias exists and is unknown inmagnitude, the true value or cannot be approximated despiteincreased replication. See random and bias deviations inA1.2.5and A1.2.6. See also Annex A2.3.4.2 outlier, na member of a set of values which isinconsistent with the other members of that set.3.4.3 reference value, na value (usually a mean) gener-ated by a recognized
