Choosing the best model in fuzzy Bayesian statistics and its application in financial analysis

Abstract

Analysts generally use closing price and normal distribution assumption for a model’s distribution in financial analysis. However, stock price fluctuation is reflected by a set of four values, namely opening, highest, lowest and closing prices. We therefore include the highest and the lowest prices to take into account more information in the hope of ending up with a more exact result as data contains a ranges of values instead of one only (i.e. the data is a form of fuzzy number). Moreover, the assumption that data is normally distributed is not always satisfied and Jacque Bera or Chi square tests are often employed to test the data’s normality. The tests require the use of pvalue which is quite controversial at present. This paper employs fuzzy Bayes point estimator to choose the most suitable distribution. On a sample of 9 stocks with large capitalization in Vietnam from their listed dates until November 06, 2015, we found that some stocks have prices distributed more reasonably than normal distribution and some are not.