Using Bayesian Statistics in Enterprise Demography Cover Image

Using Bayesian Statistics in Enterprise Demography
Using Bayesian Statistics in Enterprise Demography

Author(s): Eva Kotlebová, Ivan Láska
Subject(s): Economy
Published by: Wydawnictwo Uniwersytetu Ekonomicznego w Krakowie
Keywords: Bayesian point estimation; mean square error; conjugate family; prior distribution; posterior distribution; number of enterprise births

Summary/Abstract: Knowledge of the number of different kinds of enterprises that will be created in a coming year is essential information. It can be used in macroeconomic analyses and as a constituent of the background for economic policy.From a demographics point of view, we consider the creation (birth) of some enterprise as a basic indicator. It can also be approached from the point of view of inference, as the creation of enterprise is influenced by a wide variety of inputs. Enterprise creation may therefore be thought of as a random process.The analytic tools Bayesian statistics provide make it possible involve more kinds of information into statistical analysis and gradually update the parameter estimations. We used the conjugate family Poisson / gamma to estimate the number of enterprises to be created in a coming year. The considerations were concerned with the mean square error, which was used as the main criterion of the point estimation quality. We solved two kinds of problems: to find a Bayesian point estimation that has a smaller mean square error than the classical one in a predetermined interval, and, along with it, to model prior information in a very simple way.In finding some connection among the variables contained in the conjugate family Poisson / gamma, we solved both presented problems and also developed a simple algorithm for optimal point estimation of the Poisson distribution parameter. This algorithm was used to estimate the number of enterprises created.

  • Issue Year: 947/2015
  • Issue No: 11
  • Page Range: 67-81
  • Page Count: 15
  • Language: English