Statistisk tidsserieanalys av skillnader i partikelhalter mellan en korsning och en väglänk i Stockholm : En modellering av felterm med GARCH och Cochrane-Orcutt

Sammanfattning: In Sweden, the assumption is that there is no difference in particle halt between an intersection and a road link and this assumption is mentioned among government institutions. The intrest to investigate if statistical differences in particle halt between an intersection and a road link arises when enviromental questions concerning intersections has become more interesting.Data that is used in the thesis is between the 1:th of March 2016 at 13:15 to the 1:th of May 2016 at 00:00 and is measured at three measuring stations that are closely positioned.The stations that measure the particle halt is on the intersection at Hornsgatan 104 and the road link at Hornsgatan 110 where data of the particles with dimensions 1, 2.5 and 10 micrometer in diameter is measured. Meteorological variables is measured at Torkel Knutssonsgata and the amount of traffc is measured at Hornsgatan 110 and indicates if there is a high or low amount of traffic.The variables that are used to model the difference are produced by subtracting data from the road link of the data from the intersection. A difference between the intersection and the road link is made and modeling of the difference is possible. One difference is made for each of the particles and the differences has a variance that is not constant and sudden extreme values can appear. Time serie regressions are used for modeling the three particles (< 1, 1-2.5 and 2.5-10 micrometer in diameter). Methods such as GARCH and Cochrane-Orcutt are used in the error term where autocorrelation and heteroscedasticity are needed to be taken into account.For particles with dimensions < 1 and 1-2.5 micrometers in diameter, explained by different time variables, there is a variation of the difference over time and a signicant difference between the intersection and the road link exists. For particles with dimensions 2.5-10 micrometer in diameter there is a statistical difference in the month of April. The models with serveral explanatory variables shows that there is a number of signicant variables that can explain the general difference for each particle halt. Those variables are relative humidity, maximum wind and the direction of the wind.