Beyond the P Value

Beyond the P Value

Bayesian Statistics

Have you ever wondered how Google, Facebook and Amazon tailor adverts to you? or how pharmaceutical companies decide how many people should test the real drug or the placebo? or how the weather is predicted? or even how the Enigma code was cracked? well Bayesian Statistics is crucial for all these tasks.

Bayesian theorem uses the prior knowledge that you have about something to influence decisions. For example, the adverts you see that are tailored to you on Google or Facebook are identified using Bayesian statistics. Your internet searches are fed into an equation that calculates the most suitable adverts for you, based on your previous searches. These calculations are continually updated so you receive adverts most relevant to your recent internet searches. This is similar to the way our brains work; we constantly assimilate information and use that to influence our decisions.

Although Bayesian statistics has been successfully used in various industries for many years, it is considered controversial by some because statisticians are supposed to be impartial and using a ‘prior belief’ about the data goes against this concept.

Despite this, using historical data combined with current test data could help put new data into the context of past research, rather than treating each test as an individual study and making subjective interpretations about the data. We often have large amounts of historical data in sensory science, particularly with panellists working on certain products over time. It is therefore worth considering whether Bayesian statistics could potentially help analysis in the sensory science arena.

Areas that could benefit from the Bayesian approach are:

  • panel performance monitoring;
  • optimising the number of consumers for a test when limited time and/or product is available;
  • understanding the consistency of consumer behaviours.

Whilst the Bayesian approach is not new, it’s use in Sensory and Consumer Science is relatively so. We believe the potential is large and very interesting to explore.