### MNL63: Appendix G: Methods for Determining Whether JAR Distributions are Similar Among Products (Chi-Square, Cochran-Mantel-Haenszel (CMH), Stuart-Maxwell, McNemar)

Fritz, Carl

*Statistical Consultant, Hillsborough, NJ*

Pages: 9 Published: Jan 2009

**Abstract**

The following methods can each be used to determine whether JAR score distributions are similar among a set of products: • Chi-square method • Cochran-Mantel-Haenszel (CMH) method [1], [2] • Stuart-Maxwell method [3] • McNemar method [4] The chi-square method and CMH method are the most general of the four methods listed above. Both methods can be used for comparing JAR score distributions among any number of products for any number of JAR scale categories. The chi-square method differs from the other three methods with respect to the design of the consumer test for which it can be used. The use of the chi-square method requires independence among the assessors' responses. This limits the use of the chi-square method to situations where different groups of assessors evaluate each product. The CMH, Stuart-Maxwell, and McNemar methods are all appropriate when the assessors evaluate all of the products. These methods provide an increased level of power over the chi-square method by taking advantage of the positive correlation that typically exists among an individual assessor's ratings on two or more products evaluated during the same consumer test.

**Paper ID:** MNL11488M

**Committee/Subcommittee:** E18.03

**DOI:** 10.1520/MNL11488M

ASTM International is a member of CrossRef.

**ISBN10:** 0-8031-7010-6

**ISBN13:** 978-0-8031-7010-0