A Bayesian model for online customer reviews data in tourism research : a robust analysis

The Bayesian approach to data analysis is useful when the variables considered are already subjective or abstract, as is the case with online consumer reviews and ratings in tourism research. The Bayesian framework provides a method for combining observed data from prominent e-commerce platforms with other prior information, such as expert knowledge. Also, Bayesian statistical modelling has several advantages when the sample size of observed data is small. However, a source of uncertainty is introduced into the analysis by eliciting a unique prior distribution that adequately represents the expert's judgement. We focus on the problem in a formal Bayesian robustness context by assuming that the hospitality manager is unable to choose a functional form for the prior distribution but that he or she may be able to restrict the possible priors to a class that is suitable for quantifying the practitioner's uncertainty. Our interest is: We propose a new distribution that is suitable for fitting the rating data.We have shown how the practitioner can introduce his judgements about the feeling parameter using an appropriate prior distribution andWe develop a Bayesian robust methodology to manage hospitality managers' uncertainty using a class of prior distributions suitable for quantifying the practitioner's uncertainty. We propose a new distribution that is suitable for fitting the rating data. We have shown how the practitioner can introduce his judgements about the feeling parameter using an appropriate prior distribution and We develop a Bayesian robust methodology to manage hospitality managers' uncertainty using a class of prior distributions suitable for quantifying the practitioner's uncertainty. These ideas were illustrated using real data. We demonstrate that the Bayesian robustness methodology proposed allows us to manage this uncertainty in our model by using classes of prior distributions and how the measures of interest are transformed into intervals of interest that will allow the manager to make decisions.