Bayesian approach for estimation of harmonic oscilations with fluctuating amplitude

  • Калоян Валентинов Харалампиев 1113 София, бул. "Цариградско шосе" 125, бл. 4, каб- 413
Keywords: Bayesian spectrum analysis, harmonic component, constant amplitude, fluctuating amplitude, consumer price indices

Abstract

Abstract: This article examines cyclic components with fluctuating amplitude, where harmonic models are also used for description of the fluctuating amplitude. It is shown that a harmonic component with fluctuating amplitude can be represented as a sum of two harmonic components with constant amplitude. But since the overall model is a sum of several harmonic components with a constant amplitude, the question arises how to understand which pairs of harmonic components are simply a sum of two harmonic components with a constant amplitude and which pairs form harmonic components with fluctuating amplitude. The first step in answering this question is to evaluate a model that is not a sum of separate harmonic components but a sum of pairs of harmonic components. But before that, the article shows how to evaluate any model. The second step is to define a rule that takes the form of a system of two inequalities. Only when both inequalities are satisfied only then we can argue that there is a cyclic component with a fluctuating amplitude. For illustration, the dynamics of the price level in Bulgaria is examined. There are revealed 4 harmonic components with a constant amplitude describing the trend, and 12 harmonic components that describe cycles. Of these, 3 are with fluctuating amplitude and 9 are with constant amplitude. The resulting model has a very good explanatory power – 99.8% of the variation of the price level is explained by the obtained model.

References

1. Bretthorst, G., 1988. Bayesian Spectrum Analysis and Parameter Estimation. Berlin, Heidelberg: Springer-Verlag

2. Bretthorst, G., 1990. An Introduction to Parameter Estimation Using Bayesian Probability Theory, in P. Fougere (ed.), Maximum Entropy and Bayesian Methods, Dartmouth: Kluwer Academic Publishers, 1-27

3. Gatev, K., Koseva, D. and Spasov, A., 1991. Obshta teoriya na statistikata. Sofia: Nauka i izkustvo

4. Naydenov, G. and Haralampiev, K., 2014. Dulgite vulni na Kondratiev pri “sotsializma”. Sofia: Izdatelski kompleks – UNSS
Published
2020-12-07