Towards a model and methodology for evaluating data quality in software engineering experiments