MNL63

    Appendix U: Regression

    Published: Jan 2009

      Format Pages Price  
    PDF Version (96K) 2 $25   ADD TO CART
    Complete Source PDF (11M) 2 $55   ADD TO CART


    Abstract

    Many of the other techniques in this guideline treat the effect of JAR scales on the overall rating one scale at a time. Regression analysis allows the researcher to evaluate the joint effects of the scales levels on overall response. The stronger the relationship between a JAR scale and the overall response, the more important that “Just-Right” attribute is in explaining the liking attribute, even after controlling for the other attributes. The regression can be either non-parametric (ordinal) or parametric (ordinary regression) and the JAR scales can have either a linear or non-linear effect on the response. The examples below use linear regression because of ease of use and widespread availability in many statistical packages. The more general approaches require more statistical sophistication. Regression analysis can be done for the entire data set (all samples combined) or for each individual sample. Conducting the analysis on all samples combined gives a general overview of how the “Just Right” attributes work together to explain liking. This analysis can be conducted using either the individual respondent data or product mean scores.


    Author Information:

    Herskovic, Joseph E.
    Sensory ConAgra Foods, Inc., Omaha, NE


    Paper ID: MNL11502M

    Committee/Subcommittee: E18.03

    DOI: 10.1520/MNL11502M


    CrossRef ASTM International is a member of CrossRef.

    ISBN10: 0-8031-7010-6
    ISBN13: 978-0-8031-7010-0