1、November 2009DEUTSCHE NORM English price group 9No part of this standard may be reproduced without prior permission ofDIN Deutsches Institut fr Normung e. V., Berlin. Beuth Verlag GmbH, 10772 Berlin, Germany,has the exclusive right of sale for German Standards (DIN-Normen).ICS 67.240!$ZW“1555204www.
2、din.deDDIN ISO 16820Sensory analysis Methodology Sequential analysis (ISO 16820:2004)English version of DIN ISO 16820:2009-11Sensorische Analyse Prfverfahren Sequentialanalyse (ISO 16820:2004)Englische Fassung DIN ISO 16820:2009-11www.beuth.deDocument comprises 13 pages2 Contents Page National forew
3、ord 3 National Annex NA (informative) Bibliography .3 1 Scope 4 2 Normative references 4 3 Terms, definitions and symbols.4 4 Principle5 5 Procedure .5 DIN ISO 16820:2009-11 Annex A (informative) Examples .7 Bibliography. 13 DIN ISO 16820:2009-11 3 National foreword This standard has been prepared b
4、y Technical Committee ISO/TC 34 “Food products” (Secretariat: AFNOR, France), Subcommittee SC 12 “Sensory analysis” (Secretariat: IRAM, Argentina). The responsible German body involved in its preparation was the Normenausschuss Lebensmittel und landwirtschaftliche Produkte (Food and Agricultural Pro
5、ducts Standards Committee), Technical Committee NA 057-01-01 AA Sensorik. The original ISO document erroneously contains a dated reference to ISO 5492. This has been corrected in the German version, and correspondingly in the English version. At the time of publication of this standard ISO 5492:2008
6、 was valid. The DIN Standards corresponding to the International Standards referred to in this document are as follows: ISO 3534-1 E DIN ISO 3534-1 ISO 4120 DIN EN ISO 4120 ISO 5492 DIN EN ISO 5492 ISO 5495 DIN EN ISO 5495 ISO 6658 DIN 10950-2 ISO 8586-1 DIN 10961 ISO 10399 E DIN EN ISO 10399 Nation
7、al Annex NA (informative) Bibliography DIN 10950-2, Sensory analysis Part 2: Basic principles DIN 10961, Training of assessors for sensory analysis DIN EN ISO 4120, Sensory analysis Methodology Triangle test DIN EN ISO 5492, Sensory analysis Vocabulary DIN EN ISO 5495, Sensory analysis Methodology P
8、aired comparison test E DIN EN ISO 10399, Sensory analysis Methodology Duo-trio test E DIN ISO 3534-1, Statistics Vocabulary and symbols Part 1: General statistical terms and terms used in probability Sensory analysis Methodology Sequential analysis 1 Scope This International Standard describes a pr
9、ocedure for statistically analysing data from forced-choice sensory discrimination tests, such as the Triangle, Duo-Trio, 3-AFC, 2-AFC, in which after every trial of the discrimination test the decision can be made to stop testing and declare a difference, to stop testing and declare no difference,
10、or to continue testing. The sequential method often allows for a decision to be made after fewer trials of the discrimination test than would be required by conventional approaches that use predetermined numbers of assessments. The method is effective for a) determining that either a perceptible dif
11、ference results, or a perceptible difference does not result when, for example, a change is made in ingredients, processing, packaging, handling or storage; b) or for selecting, training and monitoring assessors. 2 Normative references The following referenced documents are indispensable for the app
12、lication of this document. For dated references, only the edition cited applies. For undated references, the latest edition of the referenced document (including any amendments) applies. 3 Terms, definitions and symbols 3.1 Terms and definitions For the purposes of this document, the terms and defin
13、itions given in ISO 5492 and the following apply. 3.1.1 alpha-risk -risk probability of concluding that a perceptible difference exists when one does not NOTE This is also known as Type I error, significance level or false positive rate. 3.1.2 beta-risk -risk probability of concluding that no percep
14、tible difference exists when one does NOTE This is also known as Type II error or false negative rate. ISO 5492, Sensory analysis Vocabulary 4 DIN ISO 16820:2009-11 3.1.3 sensitivity general term used to summarize the performance characteristics of the test NOTE In statistical terms, the sensitivity
15、 of the test is defined by the values of , and pd. 3.2 Symbols p0 probability of a correct response when no perceptible difference exists pd proportion of assessments in which a perceptible difference is detected between the two products p1 probability of a correct response when a perceptible differ
16、ence does exist 4 Principle The type of discrimination test (triangle, duo-trio, etc.) is chosen. The sensitivity of the test is defined by selecting values for , and pd. The boundaries of the decision regions are computed based on , , p0and p1. After every trial of the discrimination test, the tota
17、l number of correct responses for the panel, see Clause 1a), or per assessor, see Clause 1b) is compared to the decision boundaries to determine if testing can be stopped and a difference can be declared, if testing can be stopped and no difference can be declared, or if testing should continue. 5 P
18、rocedure 5.1 Construct a graph, as in Figure A.1, which illustrates the boundaries of the decision regions based on , , p0and p1as follows. a) and are chosen based on the risks the researcher is willing to take of obtaining a false positive or a false negative result, respectively. is the probabilit
19、y of declaring that a difference exists when the true probability of a correct response is p0. is the probability of failing to declare that a difference exists when the true probability of a correct response is p1(p1 p0). b) p0is the probability of a correct response when no perceptible difference
20、exists (i.e. the probability of a correct guess). The value of p0depends on the discrimination test being used: for the triangle and the 3-AFC tests, 01/ 3p = for the duo-trio and the 2-AFC tests, 01/ 2p = c) p1is the probability of a correct response when a perceptible difference does exist. The va
21、lue of p1depends on pd: for the triangle and 3-AFC tests, d1d13ppp=+ for the duo-trio and 2-AFC tests, d1d12ppp=+5 DIN ISO 16820:2009-11 d) the lines that form the boundaries of the decision regions are calculated as: lower line: () ( ) ( ) ( )() () ( ) ( )10010 1 0lg lg 1 lg 1 lg 1lg lg lg 1 lg 1 n
22、pnpdpp p p+=+upper line: ()() ( ) ( )() () ( ) ( )10110 1 0lg 1 lg lg 1 lg 1lg lg lg 1 lg 1 npnpdpp p p +=+where , , p0and p1are as defined above, and n is the number of trials of the test. NOTE The distance between the two lines depends on p1 p0. 5.2 After each trial of the discrimination test, plo
23、t the total number of correct responses (on the vertical axis) versus the number of trials (on the horizontal axis): if the total number of correct responses falls between the lower and upper lines on the chart, then continue testing by conducting another trial; if the total number of correct respon
24、ses falls ab ove the upper line on the chart, then stop testing and conclude that a perceptible difference exists (at the -level of significance); if the total number of correct responses falls bel ow the lower line on the chart, then stop testing and conclude that no meaningful difference exists i.
25、e. there is less than a N1)probability that the true probability of a correct response is as high as p1. N1)National footnote: In ISO 16820:2004, it is erroneously given as “(1 )”. 6 DIN ISO 16820:2009-11 Annex A (informative) Examples A.1 Example 1 Sequential analysis of a series of triangle tests:
26、 acceptance vs. rejection of two trainees on a panel A.1.1 Background A sensory analyst wishes to base the decision to accept or reject two trainees on the panel on their performance in triangle tests using a typical pair of products. Each trainee receives a series of triangle tests. Intervals betwe
27、en tests are kept long enough to avoid sensory fatigue. A.1.2 Test design The number of trials required to accept or reject a trainee is determined by sequential analysis using a graph as shown in Figure A.1. To position the boundaries of the decision regions (i.e., the two lines in Figure A.1), ass
28、ign a value to each of the four parameters, , , p0and p1. In the triangle test 01/ 3p = (i.e. the probability of a correct guess, pd= 0). Usually the minimum acceptable rate of detection is set at pd= 50 %, which makes 1120,50 (1 0,50) ( )33p =+ = If it is desired to reduce the number of trials to r
29、each a decision, lower the minimum acceptable rate of detection e.g. to pd= 40 %, which makes 110,40 (1 0,40) ( ) 0,603p =+ =, etc. NOTE In this example the definition of pdis not the proportion of the population of assessors who can distinguish the samples but rather the proportion of trials in whi
30、ch a single assessor actually distinguishes the samples. The analyst chooses the following values for the parameters: = 0,05 is the probability of selecting an unacceptable trainee; = 0,10 is the probability of rejecting an acceptable trainee; 01/ 3p = is the maximum unacceptable ability (i.e. the n
31、ull hypothesis p-value of the triangle test); 12/3p = is the minimum acceptable ability (i.e. the probability that the odd sample will be detected when pd= 0,50). 7 DIN ISO 16820:2009-11 A.1.3 Analysis and interpretation of results As each triangle is completed, the results are entered in the diagra
32、m in Figure A.1 as follows. Enter the result of the first trial, if correct, as (x, y) = (1,1) and, if incorrect, as (x, y) = (1,0). For each succeeding trial, increase the value of x by 1 and increase y by 1 for a correct response, or increase x by 1 and y by 0 for an incorrect response. Continue t
33、esting until a plotted point touches or crosses either of the decision boundaries. Draw the indicated conclusion (i.e. accept or reject the trainee). Trainee A is correct in all tests and is accepted after five trials. Trainee B fails in the first triangle, succeeds in triangles 2 and 3, but then fa
34、ils on every subsequent triangle and is rejected after the 8thtrial. = 0,05 = 0,10 Parameters of the test: p0= 13p1= 231001100lg( ) lg(1 ) lg(1 ) lg(1 )Lower:lg( ) lg( ) lg(1 ) lg(1 ) npnpdpp p p+=+0lg(0,10) lg(1 0,05) lg 1 (2/3) lg 1 (1/3)Lower:lg(2/3) lg(1/3) lg 1 (2/3) lg 1 (1/3)nnd + =+0Lower: 1
35、,624 0,5dn= + Boundary lines: 10110 1 0lg(1 ) lg( ) lg(1 ) lg(1 )Upper:lg( ) lg( ) lg(1 ) lg(1 ) npnpdpp p p + =+1lg(1 0,10) lg(0,05) lg 1 (2/3) lg 1 (1/3)Upper:lg(2/3) lg(1/3) lg 1 (2/3) lg 1 (1/3)nnd + =+1Upper: 2,085 0,5dn=+ 8 DIN ISO 16820:2009-11 Key Trainee A Trainee B n is the number of trial
36、s d is the number of correct responses A is the acceptance region B is the continue testing region C is the rejection region region NOTE At the 5thtrial, Trainee A passes out of the “continue testing region” and into the “acceptance region”. At the 8thtrial, Trainee B passes out of the “continue tes
37、ting region” and into the “rejection region”. Figure A.1 Use of sequential analysis in triangle tests Example 1: Selection of two trainees 9 DIN ISO 16820:2009-11 A.2 Example 2 Sequential analysis of a series of duo-trio tests: Warmed-over flavour (WOF) in beef patties during storage A.2.1 Backgroun
38、d A manufacturers quality control panel has detected warmed-over flavour (WOF) in beef patties refrigerated for 5 days and then reheated. The project leader wants to set a reasonable maximum for the number of days the beef patties may be refrigerated. A.2.2 Test design Preliminary evaluations have s
39、hown that beef patties stored for 5 days exhibit strong WOF while patties stored for 1 day exhibit no WOF. The sensory analyst chooses to run a series of duo-trio tests involving patties that have been stored for 1, 3 and 5 days. Each of the stored patties will be compared to a freshly grilled contr
40、ol sample that has undergone no refrigerated storage. The three sample pairs (control vs. 1-day storage, control vs. 3-day storage and control vs. 5-day storage) are presented in separate duo-trio tests. The freshly grilled control is used as a constant reference in all three tests. As each assessor
41、 completes an evaluation, the result is added to the previous responses and the cumulative number of correct responses is plotted as illustrated in Figure A.2. The test series continues until the stored sample is declared either similar to or different from the freshly grilled control. The analyst c
42、hooses the following values for the parameters: = 0,10 is the probability of accepting an unacceptable patty; = 0,10 is the probability of rejecting an acceptable patty; 00,50p = is the probability of a correct response when no perceptible difference exists (i.e. the null hypothesis p-value of the d
43、uo-trio test); 10,70p = is the probability that the stored sample is selected as being different from the control when pd= 0,40 (i.e. p1= 0,40 + (1 0,40) (0,50) = 0,70). A.2.3 Analysis and interpretation of results Given the analysts choices for , , p0and p1, the equations for the lines that form th
44、e boundaries of the decision regions, presented in Clause 5, are: d0= 2,59 + 0,60 n and d1= 2,59 + 0,60 n. The lines are plotted in Figure A.2 along with the cumulative number of correct responses (see Table A.1) from the tests on the three stored samples. = 0,10 = 0,10 Parameters of the test: p0= 0
45、,50 p1= 0,70 10010 1 0lg( ) lg(1 ) lg(1 ) lg(1 )Lower:lg( ) lg( ) lg(1 ) lg(1 ) npnpdpp p p+=+0lg(0,10) lg(1 0,10) lg(1 0,70) lg(1 0,50)Lower:lg(0,70) lg(0,50) lg(1 0,70) lg(1 0,50)nnd += 0Lower: 2,59 0,60dn= + Boundary lines: 1011010lg(1 ) lg( ) lg(1 ) lg(1 )Upper:lg( ) lg( ) lg(1 ) lg(1 ) npnpdpp
46、p p + =+1lg(1 0,10) lg(0,10) lg(1 0,70) lg(1 0,50)Upper:lg(0,70) lg(0,50) lg(1 0,70) lg(1 0,50)nnd += 1Upper: 2,59 0,60dn=+ 10 DIN ISO 16820:2009-11 Table A.1 Results obtained in Example 2: Sequential analysis of a series of duo-trio tests Warmed-over flavour (WOF) in beef patties during storage Tes
47、t A Test B Test C Assessor (1-Day) (3-Day) (5-Day) Results Count Results Count Results Count 1 I 0 I 0 C 1 2 I 0 C 1 C 2 3 I 0 I 1 C 3 4 C 1 C 2 C 4 5 I 1 I 2 I 4 6 C 2 C 3 C 5 7 I 2 I 3 C 6 8 C 3 C 4 C 7 9 I 3 C 5 I 7 10 C 4 C 6 C 8 11 I 4 C 7 C 9 12 I 7 C 10 13 C 8 14 C 9 15 C 10 16 C 11 17 I 11 1
48、8 I 11 19 C 12 20 C 13 21 I 13 22 I 13 23 I 13 24 C 14 25 I 14 26 C 15 27 C 16 28 C 17 29 C 18 30 C 19 NOTE Results: I = incorrect, C = correct; Count = cumulative number of correct results. 11 DIN ISO 16820:2009-11 The sample stored for 1 day is declared to be similar to the freshly grilled control after 11 trials of the duo-trio test. The sample stored for 5 days is declared to be different from the control after 12 trials. The sample stored for 3 days cannot be declared similar to, or different from, the control after 30 trials (see Figure
