Abstract:

 

A measuring instrument can consistently report inaccurate information. When this occurs, the instrument is biased. All measuring instruments are prone to various types of biases. An analytical balance is calibrated to eliminate bias so that the data generated can be relied on. Where do biases occur in sensory and consumer science? Can we identify sources of bias, estimate their influence on the data and correct for them, or even better, eliminate them? Answering these questions is needed if we want to collect experimental data that translates into reliable and actionable business decisions.

In this webinar, we will review areas of sensory and consumer research where biases can typically be found, illustrate their effects, and propose solutions to reduce or even eliminate their influence on a study's outcomes. We will consider:

Position and Sequence Effects:  We will discuss the strengths and weaknesses of random rotations, Latin squares in general, Williams Squares, and an approach that optimizes position, sequence and sequence spread in a study design. Of special interest is the choice of method that can minimize bias and decrease variance for multiple-day studies and incomplete block arrangements.

Response Bias:  A simple question such as "Which one of two samples do you prefer, or do you have no preference between the two?" has inherent bias since directional preference can be recorded even if the two samples are identical. We will illustrate how to use an identicality norm to extract deeper insights including diagnosing consumer segmentation.

Code Bias: We will discuss how codes themselves can introduce bias and how to avoid it.

This webinar is intended for a general audience of industrial and academic sensory professionals and graduate students. 

 

A digital recording of this webinar is available for purchase ($289).