Survey Models for a Vehicle Speed Survey

Sammanfattning: The impact of some errors associated with a national road traffic survey is examined in this thesis. The current survey aims to evaluateefforts to reduce the speed of traffic on Swedish roads; it covers both state and urban roads, though the thesis considers only urban roads.In the survey, observational sites are selected by athree-stage sampling procedure. A measurement device installed onthe road is used to collect data, from which the average speed oftraffic on the roads is estimated. This thesis focuses on errors in the frames used in the final sampling stage, and on errors dueto missing data. The impact of these errors on the total error ofthe survey estimators is investigated. Also explored arepossibilities for reducing the total error by weighting adjustments for missing data and by reallocating the sample over the threesampling stages. The problems are approached partly theoretically,by use of various error models; partly empirically,by collectingdata on the errors. Throughout, the sampling design of the survey istaken properly into account. Our conclusion is that the frame errorunder consideration probably does not bias the estimator ofaverage speed, and it only implies a minor increase of its variance.It remains unclear whether the estimator needs to be adjustedfor missing data: however, a theoretical framework forfurther investigations is provided. For unchanged total sample size,the precision of the estimator is likely to improve if the samplesizes in the third stage are increased, and the sampling sizes inthe first stage are decreased correspondingly.