Vremenski nizovi sa negativnim celobrojnom vrednostima generisani zavisnim brojačkim nizovima
Time series with non-negative integer values based on dependent counting series
Miletić Ilić, Ana V.
Faculty:University of Niš, Faculty of Science and Mathematics
Ristić, Miroslav M.
Popović, Biljana Č.
Stanković, Miomir S.
Nastić, Aleksandar S.
MetadataShow full item record
New generalized thinning operators based on Bernoulli sequences of dependent random variables are presented In this thesis. Using these operators autoregressive models of the first order with non-negative integer values and geometric marginal distribution are constructed. Also, a mixed model based on both dependent and independent Bernoulli counting series is constructed by combined application of the binomial and the generalized binomial thinning operator. Some basic features of all introduced models are determined, unknown model parameters are estimated and asymptotic properties of obtained estimates are considered. Possible application to real data is also presented, where the new models are compared with each other and with the existing relevant INAR models.
Keywords:INAR modeli; INAR models; binomni tining; generalizovanji binomni tining; zavisni Bernulijev brojački niz; geometrijska marginalna raspodela; binomial thinning; generalized binomial thinning; dependent Bernoulli counting series; geometric marginal distribution