The INLA approach to estimation of SAR models
Translated title
INLA tilgangen til estimation af SAR modeller
Author
Vestergaard, Claus Høstrup
Term
4. term
Education
Publication year
2014
Submitted on
2014-07-11
Pages
58
Abstract
Specialet undersøger, hvordan man estimerer en Spatial Autoregressive (SAR) model ved hjælp af Integrated Nested Laplace Approximations (INLA). En SAR-model beskriver, at observationer i geografisk nærliggende områder kan hænge sammen. En central del af modellen er den rumlige vægtmatrix, som angiver, hvilke steder der påvirker hinanden, og hvor stærk påvirkningen er. Afhandlingen fokuserer på, hvordan forskellige specifikationer af denne vægtmatrix påvirker de estimerede parametre, og belyser dermed, hvor følsomme resultaterne er over for dette valg.
This thesis examines how to estimate a Spatial Autoregressive (SAR) model using Integrated Nested Laplace Approximations (INLA). A SAR model captures the idea that observations in nearby locations can be related. A key component is the spatial weight matrix, which specifies which places influence each other and by how much. The thesis focuses on how different specifications of this matrix affect the estimated parameters, highlighting how sensitive the results are to this choice.
[This abstract was generated with the help of AI]
Documents