Bayesian technique for assessing parameters’ uncertainty in probability models
Bayesian technique for assessing parameters’ uncertainty in probability models

Summary/Abstract: In this study some probability models are proposed (Normal, Log-normal and Gumbel models) to predict the quarterly inflation rate in USA in 2014 and 2020, starting from the historical quarterly predictions of the Survey of Professional Forecasters (SPF). The Normal model generated the highest inflation rates in both years (values between 4.37% and 4.77% in 2014 and values around 6% in 2020). The Bayesian statistical approach was used to assess the parameters’ uncertainty, the parameters being treated as random variables. The degree of uncertainty is rather low and there are not high differences between predictions regarding the uncertainty. The combined forecasts based on those probability models have a lower degree of anticipated uncertainty compared to the initial predictions.

• Details
• Contents