Application of ordinal regression model to analyze service quality of Riga coach terminal
The considered problem has arisen as a result of collaboration with Riga Coach Terminal (Rīgas Starptautiskā Autoosta) authorities. Recent studies of the role of buses and coaches seem to confirm the already excellent safety, environmental and social record of bus and coach transport. In Latvia this mode of transport is in competition with railway (and also private cars) that's why the quality of services is very important from all points of view. In authors’ previous researches different methods were applied to estimate functional form between overall quality of service and explanatory variables included questionnaire items related to the satisfaction accessibility (availability), information, time characteristics of service, customer service, comfort, safety, infrastructure and environment at Riga Coach Terminal. Such kind of model allows estimating influence of particular quality indicators on the overall quality assessment and simplifying the monitoring of quality indicators. In the given work ordinal regression method has been used to model the relationship between the ordinal outcome variable, e.g. estimates of overall quality of service – y i (these estimates are made on the basis of (1÷5) scale), and the 22 particular attributes of quality distributed on the mentioned above 7 groups. The main decisions involved in the model building for ordinal regression determine, which particular attributes should be included in the model, and choose the link function (e.g. logit link or complementary log–log link) that demonstrates the model appropriateness. The model fitting statistics, the accuracy of the classification results, and the validity of the model assumption, e.g., parallel lines, have been assessed for selecting the best model. The model was done on the basis of results of questionnaire of transport experts, which had been fulfilled in spring 2009. In total 44 questionnaires have been returned, however some questions remained without an answer; that's why different methods of data imputation have been applied to substitute skips in dataset and few models have been constructed for selecting the best one.
First Published Online: 03 Apr 2013
This work is licensed under a Creative Commons Attribution 4.0 International License.