Dealing with spatial autocorrelation in Indonesia marine fisheries catch production model: a spatial filtering with eigenvector approach Evellin Dewi Lusiana(a,b*), Henny Pramoedyo (c)
a) Doctoral Student of Mathematics Program, Faculty of Mathematics and Natural Sciences, Universitas Brawijaya, Malang 65145, Indonesia
*evellinlusiana[at]ub.ac.id
b) Department of Aquatic Resource Management, Faculty of Fisheries and Marine Science, Universitas Brawijaya, Malang 65145, Indonesia
c) Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Brawijaya, Malang 65145, Indonesia
Abstract
Indonesia is an archipelago country whereas 70% of its region consists of water. Therefore, the fisheries sector has great potency to support the economic growth of the country. Moreover, capture fisheries contributed as the main export commodities in Indonesia. Each province in Indonesia has its own marine area, thus the marine fisheries capture production is available in province-level data. The total production of marine fisheries capture is highly affected by its production factors such as the number of fisheries households and ships. This relationship can be modeled using a simple model like linear regression. However, since the data is considered for having spatial connectivity, then the linear regression model will suffer spatially autocorrelated residuals which violent the model assumption. The goal of this research was to deal with spatial autocorrelation in linear regression to improve the inference and validity of the model by using spatial filtering with the eigenvector (SFE) approach. The results showed that the linear regression model with eigenvector variable addition has no significant spatial autocorrelation and no heteroscedasticity.
