Analyzing the UK tourist outbound market using time series analysis, generalized linear, gee and multilevel models

Abstract

The tourism industry is affected by many factors and one of these is climate. The aim of this dissertation is to relate the number of tourists travelling out of the UK and the average Air Malta flight fare during 2010 with season, flight status and climatic variables including daily average air and sea temperatures, relative humidity, average wind speed and hours of bright sunshine averaged across a six-year period from year 2005 to year 2010. Time Series analysis is initially carried out to identify autocorrelation and trend components within the data that should be accounted for. After differencing the number of tourists travelling from UK with Air Malta and the average daily fare per flight, three models are used to relate the differenced responses to the explanatory variables. Generalized linear models (GLM) relate a response variable to the linear predictor through any invertible link function assuming that the responses are independent and the error distribution is a member of the exponential family. Generalized Estimating Equations (GEE) models, which are more appropriate for analyzing longitudinal and clustered data, extend GLMs by relaxing the assumption of independence. This is carried out by identifying an appropriate correlation structure for the data. Multilevel models, also known as hierarchical linear models, are used to accommodate the hierarchical nested structure within the data. Unlike GLM and GEE models the equations defining the hierarchical linear model contains an error term for each level of nesting.