Bayesian statistics (BUGS) in the estimation of the econometric model of grain yield in Estonian counties
2001
Päldaru, R. | Roots, J. (Estonian Agricultural University, Tartu (Estonia). Inst. of Informatics)
The authors have constructed a special regression model (econometric model) to explain the relationship between the countylevel grain yield in Estonian counties and 12 explanatory variables. The data what was used was a panel of fifteen Estonian counties observed during the period from 1994 to 1999. The most widely used methods for extracting information from the data - the ordinary linear regression (OLR) and the principal component regression (PCR) gave inadequate estimates for the parameters of the econometric model of grain yield. In this paper the authors consider the possibilities of the Bayesian regression (BR) for estimating the parameters of the econometric model of grain yield and discuss the possibilities of the Bayesian methods for estimating regression parameters. The general-purpose MCMC-based software, BUGS was used for estimating the parameters of the econometric model. An acceptable econometric model of grain yield was obtained when the informative priors were assigned for the parameters for the nutrients. The estimates of the parameters of the econometric model do not depend on the parameters of the informative prior for most variables with non-informative prior, whereas the BUGS estimates and OLR estimates for the variables with noninformative prior do not differ substantially. The plot of the BUGS estimates depending on the logarithm of the standard deviation of the prior distribution was used for assigning precision to the prior distribution. The Bayesian regression (BR) for estimating the parameters of econometric model of grain yield gave acceptable estimates for the model parameters and may be recommended for that use
显示更多 [+] 显示较少 [-]