ASTM F2340-2005(2016) Standard Specification for Developing and Validating Prediction Equation(s) or Model(s) Used in Connection with Livestock Meat and Poultry Evaluation Device(s.pdf

《ASTM F2340-2005(2016) Standard Specification for Developing and Validating Prediction Equation(s) or Model(s) Used in Connection with Livestock Meat and Poultry Evaluation Device(s.pdf》由会员分享，可在线阅读，更多相关《ASTM F2340-2005(2016) Standard Specification for Developing and Validating Prediction Equation(s) or Model(s) Used in Connection with Livestock Meat and Poultry Evaluation Device(s.pdf（3页珍藏版）》请在麦多课文档分享上搜索。

1、Designation: F2340 05 (Reapproved 2016)Standard Specification forDeveloping and Validating Prediction Equation(s) orModel(s) Used in Connection with Livestock, Meat, andPoultry Evaluation Device(s) or System(s) to DetermineValue1This standard is issued under the fixed designation F2340; the number i

2、mmediately following the designation indicates the year oforiginal adoption or, in the case of revision, the year of last revision. A number in parentheses indicates the year of last reapproval. Asuperscript epsilon () indicates an editorial change since the last revision or reapproval.1. Scope1.1 T

3、his specification covers methods to collect and analyzedata, document the results, and make predictions by anyobjective method for any characteristic used to determine valuein any species using livestock, meat, and poultry evaluationdevices or systems.1.2 This standard does not purport to address al

4、l of thesafety concerns, if any, associated with its use. It is theresponsibility of the user of this standard to establish appro-priate safety and health practices and determine the applica-bility of regulatory requirements prior to use.2. Referenced Documents2.1 ASTM Standards:2F2463 Terminology f

5、or Livestock, Meat, and Poultry Evalu-ation Systems3. Terminology3.1 For definitions of terms used in this specification, referto Terminology F2463.3.2 Definitions of Terms Specific to This Standard:3.2.1 accuracy, nstatement of the exactness with which ameasurement approaches the true measure for t

6、hat character-istic; accuracy is contrasted with precision, which is concernedwith the repeatability of the measurements. Therefore, with alarge bias, a measurement may be of high precision, but of lowaccuracy.3.2.2 calibration data set, ndata set used to develop theinitial prediction equations; sam

7、e as developmental or predic-tion data set.3.2.3 coeffcient of determination, npercentage of variabil-ity in the response (dependent) variable that can be explainedby the prediction equation.R25 1 2(y 2 y!2(y 2 y!23.2.4 root mean square error for calibration, nsquare rootof the sum of squared residu

8、als divided by nc(k + 1), wherencis the sample size for the calibration data set, and k is thenumber of explanatory variables in the prediction equation.(y 2 y!2nc2 k11!3.2.5 root mean square error for validation, nsquare rootof the sum of squared residuals divided by ny, where nyis thesample size f

9、or the validation data set.(y 2 y!2nv3.2.6 validation data set, nthe data set used to test thepredictive accuracy of the equations developed from thecalibration data set.3.2.7 value, commerce, nmeasure of economic worth incommerce.4. Significance and Use4.1 The procedures in this specification are t

10、o be used by allparties interested in predicting composition or quality, or both,for the purpose of establishing value based upon device orsystem measurements. Whenever new prediction equations areestablished, or when a change is experienced that could affectthe performance of existing equations, th

11、ese procedures shallbe used.5. Procedure5.1 Experimental Design:5.1.1 Define the Population for Development of a PredictionEquation:5.1.1.1 To establish the predictive ability and validity of anequation(s) using measures (independent variables) from an1This specification is under the jurisdiction of

12、 ASTM Committee F10 onLivestock, Meat, and Poultry Evaluation Systems and is the direct responsibility ofSubcommittee F10.40 on Predictive Accuracy.Current edition approved Sept. 1, 2016. Published September 2016. Originallyapproved in 2004. Last previous edition approved in 2010 as F2340 05 (2010).

13、DOI: 10.1520/F2340-05R16.2For referenced ASTM standards, visit the ASTM website, www.astm.org, orcontact ASTM Customer Service at serviceastm.org. For Annual Book of ASTMStandards volume information, refer to the standards Document Summary page onthe ASTM website.Copyright ASTM International, 100 Ba

14、rr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. United States1evaluation device or system, it is necessary to define thepopulation on which the prediction model is intended to beused.(1) The species on which measurements will be made mustbe defined.(2) The population for scope of use

15、 must be clearly defined.This may include, but is not limited to, factors such asgeographical location, gender, age, breed type, or any otherfactor that may affect the equation accuracy.(3) The characteristic to be predicted must be clearlydefined.5.1.2 Select a Sample Population for Development of

16、aPrediction Equation:5.1.2.1 The sample size for the calibration data set must beat a minimum 10k, where k is the number of variables in theprediction equation, or 100 observations, whichever is greater.The sample size for the validation data set must be at least20 % of the size of the calibration v

17、alidation data set. Forexample, if the prediction equation has five explanatoryvariables, the calibration data set will require a minimum of100 observations and the validation set must have at least 20observations. These are minimal requirements; larger samplesizes are encouraged, keeping in mind th

18、at the calibration dataset must be larger than the validation data set.5.1.2.2 The sample size must be large enough to be repre-sentative of the population; otherwise the resultant equationwill not be suitable for use in the population to which theequation will be applied. This may require a larger

19、sample sizethan the minimal requirement in 5.1.2.1. When possible, it maybe useful to refer to existing data sets that describe a particularpopulation to ensure that the sample includes most of thevariation in the population. For example, if one were develop-ing an equation to predict yield grade in

20、 U.S. fed beef packingplants, one would want to make sure that the samples used todevelop and validate the regression model encompassed mostof the normal variation in yield grade, yield grade factors, andfactors that might affect the accuracy of the model. In thisexample, the simple statistics of th

21、ese characteristics in thecalibration data sets should be compared to the simple statisticsof these characteristics in references such as the National BeefQuality Audits. Users are encouraged to work with a statisti-cian.5.1.3 Develop an Experimental ProcessA clearly definedprocess must be establish

22、ed and documented. That process,which includes consistent, repeatable methods, should be usedto obtain the measurements under the same conditions in whichthe device or system would be expected to operate. Inparticular, the validity of the approach and the repeatability ofthe procedure must be docume

23、nted and demonstrated. Formany of the common characteristics to be predicted (such aspercent lean), there are a number of reference methods com-monly accepted within the discipline. Where accepted methodsexist, they should be used and cited. Where accepted methodsdo not exist, a sound, science-based

24、 process of methoddevelopment should be followed. Consideration should begiven to sources of variation for the measurements andstrategies to minimize any bias that may exist.5.1.4 Independent Third-Party ConsultationAfter the ex-perimental process has been established (but before initiationof the sa

25、mpling), it is recommended that the users obtain anindependent third-party consultation to review the proceduresfor compliance with the guidelines established in the previoussections. The consultation should focus on areas such as thenumber of samples, the sample selection protocol, and theproject p

26、rocedures to ensure that the process will allow theusers to determine effectively the predictive ability and validityof the equation or model.5.1.5 Develop the Model or Equation:5.1.5.1 Collect data for the calibration (developmental) dataset and develop the model or equation. Report the value of th

27、ecoefficient of determination, R2, for the calibration data set.5.1.5.2 Describe the sample used to develop the model orequation. Calculate the simple statistics (standard deviation,mean, minimum, and maximum values) of the data set that wasused to develop the prediction model (calibration data setf

28、orexample, see Table 1).5.1.6 Validation of Prediction Models or Equation(s):5.1.6.1 ObjectiveTo demonstrate the validity of the initialprediction model or equation with a different sample.5.1.6.2 Select a sample for validation of a prediction equa-tion.Ageneral recommendation is for the size of the

29、 validationdata set to be 20 % of the size of the calibration data set.However, the sample must be large and variable enough to berepresentative of the population to which the equation or modelwill be applied (refer to the calibration data set statistics forguidance).5.1.6.3 Collect data for the val

30、idation data set. In validationtrials, data used to determine predicted values must be col-lected under conditions where the devices or systems will beused or as close to on-line as possible. Prediction equationsmust not be applied to populations whose range of relevantcharacteristics (independent a

31、nd dependent variables or factorsTABLE 1 Simple Statistics of Beef Carcass Characteristics for Calibration Data Set (n = 400)Data Set Characteristic Mean SD Minimum MaximumCalibration Hot carcass weight, kg 351 41 227 460Calibration Marbling scoreA505 106 250 1090Calibration Preliminary yield grade

32、3.07 0.58 2.1 5.5Calibration Adjusted preliminary yield grade 3.29 0.62 2.0 5.6Calibration Adjustment of preliminary yield grade 0.22 0.22 -0.3 1.1Calibration Kidney, pelvic, and heart fat, % 2.08 0.69 0.0 4.5Calibration Longissimus area, cm290.2 11.3 53.5 135.5Calibration Yield grade 2.65 1.06 -0.5

33、 6.3A200 = Practically Devoid0; 300 = Traces0; 400 = Slight0; 500 = Small0; 600 = Modest0; 700 = Moderate0; 800 = Slightly Abundant0; 900 = Moderately Abundant0; 1000= Abundant0.F2340 05 (2016)2that affect the relationship between independent and dependentvariables) is practically different from the

34、 range of thosecharacteristics in the calibration data set.5.1.6.4 Apply the model or equation from the calibrationdata set to the validation data set and evaluate the differencebetween the predicted and actual values. Report the coefficientof determination, R2, for the validation data set.5.1.6.5 T

35、he model or equation would be deemed valid if theroot mean square error for validation is within 20 % of the rootmean square error for calibration. The model or equationwould be deemed invalid if the root mean square error fromvalidation is greater than the root mean square error fromcalibration by

36、more than 20 % of the root mean square errorfrom calibration.5.1.6.6 Describe the sample used to validate the model orequation. Calculate the simple statistics (standard deviation,mean, minimum, and maximum values) of the data set. Thesestatistics define the bounds of the population to which themode

37、l or equation can be applied.5.2 Revalidation of Prediction Model(s) or Equation(s):5.2.1 Potential Factors Influencing the Decision to Revali-date:5.2.1.1 One factor is the range of the independent variablesor the prediction of the dependent variable used in thevalidation sample is practically diff

38、erent than the range in thecurrent population (“practically different“ can be defined as tenor more percent of the population measured fall outside therange of independent variables for the validation data set; userscarry the burden of proof to demonstrate no practical changeduring the cumulative pr

39、eceding 12 month period). Thisevaluation should be conducted at least annually and shouldencompass all relevant observations (observations in which thepredicted value influenced seller payment) in the previous 12month period.5.2.1.2 A second factor is a device or system modification/change that prac

40、tically affects at least one independent variablethat causes a change that is greater than the establishedtolerance.5.2.1.3 A third factor is a process modification/change thatpractically affects at least one independent variable that causesa change that is greater than the established tolerance.5.2

41、.2 Determination of Appropriate Corrective Action:5.2.2.1 Compare differences in independent variable rangesor the predicted values to determine if a practical reason torevalidate exists.5.2.2.2 Upon identification of a practical reason torevalidate, operators must initiate revalidation promptly and

42、notify any appropriate government entity, if required.5.3 Documentation of Results:5.3.1 Records:5.3.1.1 Users must maintain written records of the data,procedures, and test results used to validate or revalidate aprediction equation; these records must be maintained so longas they are the basis for

43、 the validation of a prediction equationor model in use.5.3.1.2 Parties using prediction equations or models todetermine value of livestock, meat, or poultry have a right toprotect trade secrets, subject to applicable laws.5.3.2 RepresentationsAll representations or claims aboutthe performance of a

44、prediction equation or model must useterminology that is consistent with this standard, supported byavailable data, and based upon the procedures outlined above.6. Keywords6.1 coefficient of determination; model; prediction equation;revalidation; root mean square error; standard deviation; vali-dati

45、onASTM International takes no position respecting the validity of any patent rights asserted in connection with any item mentionedin this standard. Users of this standard are expressly advised that determination of the validity of any such patent rights, and the riskof infringement of such rights, a

46、re entirely their own responsibility.This standard is subject to revision at any time by the responsible technical committee and must be reviewed every five years andif not revised, either reapproved or withdrawn. Your comments are invited either for revision of this standard or for additional stand

47、ardsand should be addressed to ASTM International Headquarters. Your comments will receive careful consideration at a meeting of theresponsible technical committee, which you may attend. If you feel that your comments have not received a fair hearing you shouldmake your views known to the ASTM Commi

48、ttee on Standards, at the address shown below.This standard is copyrighted by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959,United States. Individual reprints (single or multiple copies) of this standard may be obtained by contacting ASTM at the aboveaddress or at 610-832-9585 (phone), 610-832-9555 (fax), or serviceastm.org (e-mail); or through the ASTM website(www.astm.org). Permission rights to photocopy the standard may also be secured from the Copyright Clearance Center, 222Rosewood Drive, Danvers, MA 01923, Tel: (978) 646-2600; http:/ 05 (2016)3