FESTA. D2.4 Data analysis and modelling
reportposted on 22.12.2009, 16:42 authored by James Lenard, FESTA Consortium, Andrew Morris
The chapter of the handbook and the deliverable on data analysis will provide guidance and general principles for - pre-testing to check the usability of the system and the feasibility of the evaluation process, - controlling the consistency of the chain and the precision with different sampling schemes, - modelling the impact for each indicators and for an integrated evaluation including a systemic and multidisciplinary interpretation of the effects, - integrating and controlling the quality of space-time data from various sources (numerical, video, questionnaires), - selecting the appropriate statistical techniques for data processing, PI estimation and hypothesis testing in accordance to the list of indicators and experimental design, - scaling up from experimental data and identified models to population and network level. Experimentalists stress the role and importance of a preliminary field test in FOT. Three main objectives have been defined to make a preliminary diagnosis of usability of the systems and to check the relevance and feasibility of the evaluation process. These preliminary tests are very important for the practical deployment of the FOT as well as for the overall scientific evaluation process. Recommendations about the monitoring of local and global consistency of the chain of operations from the database extraction to the hypothesis testing are given, especially to ensure the validation of the calculation of the Performance indicators. Integration of the outputs of the different analysis and hypothesis testing requires a kind of meta-model and the competences of a multidisciplinary evaluation team, specially for interpretation of the system impact and secondary effects (behavioural adaptation, learning process, long-term retroaction, …). In cooperation with WP2.2, methods for data quality control have been defined. Four types of checks have been defined to complement the information of the data base in order to prepare the data for the analysis. Statistical methods have been described for three steps of the chain: data processing, PI calculation and hypothesis testing. They belong either to exploratory data analysis or to inferential analysis. Special attention has been given to the precision of the estimates of the effects or impacts of the system on the Performance indicators by stressing the importance of controlled randomisation and application of mixed regression models. Scaling-up relies upon the potential to extrapolate from the PIs to estimates of the impact at an aggregated level. Three approaches have been defined to carry out the scaling up process from direct estimations to simulation models with the related assumptions. Models and methodologies for scaling up results on traffic flow, environmental effects (e.g. PM10, CO2, Noise emissions in db) and traffic safety have been collected.