4425

4175

كارشناسي ارشد

آمار اقتصادي و اجتماعي

اصفهان: دانشگاه صنعتي اصفهان، دانشكده علوم رياضي

1387

[هشت]،158،[II]ص:

ص.ع به :فارسي و انگليسي

حميد قرباني

صفيه محمودي

انگليسي به فارسي -فارسي به انگليسي

18/1/88

هوشنگ طالبي،سروش عليمرادي

1396/09/11

كتابنامه

رياضي

ID4175

به فارسي و انگليسي :قابل رويت در نسخه ديجيتال

Spatial Zero In ated Models Ahad Rahimpoor rahimpoora@math iut ac ir October 14 2008 Master of Science Thesis Department of Mathematical Sciences Isfahan University of Technology Isfahan 84156 83111 IranSupervisor Dr Hamid Ghorbani hamidghorbani@cc iut ac ir Advisor Dr Sa eh Mahmoodi mahmoodi@cc iut ac ir Department Graduate Program Coordinator Dr Rasoul Nasr Isfahani isfahani@cc iut ac ir Department of Mathematical Sciences Isfahan University of Technology Isfahan 84156 83111 Iran AbstractIn this thesis we present an expanded account of Spatial Zero In ated Models based onan article by Rathbun and Fei 2006 Ecological counts data are often characterized byan excess of zeros and spatial dependence Excess zeros can occur in regions outside therange of the distribution of a given species Furthermore their intent may be to obtain abetter understanding of what environmental factors or habitat conditions are favorable tothe species of interest In this thesis we begin with the presentation of some classes of zero in ated and parameter in ated distributions Zero in ated distributions have two forms 1 a mixture of a degen erate distribution with mass at zero and a non degenerate distribution such as the binomialor Poisson distribution and 2 a conditional speci cation where the the zero mass and thetruncated distribution of the non zero counts are modelled independently In ated parameterdistributions are an extension of the generalized power series distributions by including anadditional parameter p It has a natural interpretation in terms of zero in ation Therelationships between the in ated parameter distributions according to the remaining pa rameters are the same as between their classical analogue As an application a data setis tted by Poisson distribution zero in ated and parameter in ated Poisson distributions Zero in ated data are not uncommon yet they are not handled well by standard models Then spatial regression models and spatial zero in ated regression models are introducedfor count data For count data generalized linear models is used to model the mean of countdistribution Spatial correlation is introduced thought a latent process For zero in ateddata the excess zeros are generated with probability p and count data are generated from aPoisson or another distribution with probability 1 p Logistic regression is used to modelthe probabilities of the excess zeros while the log linear model is used for the Poisson mean Spatial dependence is introduced by adding spatially dependent random e ects to the logisticregression and or log linear models Another way for generating excess zeros is spatial probitmodel Under this model an excess zero is generated at a given site if the realization ofa normal random eld falls below a threshold Here the realization of the random eld isinterpreted to be a measure of habitat suitability Finally this models have been tted tosimulated data using R and WinBUGS softwares 1

حميد قرباني

صفيه محمودي

هوشنگ طالبي،سروش عليمرادي