A new approach for data quality analysis : in the context of the two parameter Weibull live cycle models

This thesis is concerned with the two-parameter Weibull distribution, in the context of data quality analysis for life cycle testing models. Data quality can´t be neglected while fitting such a distribution with field data. The focus of this work is on parameter estimation, in relation to sample size and confidence bounds in the presence of complete and randomly right censored data. Confidence bounds have been evaluated through simulations with two different methods, using maximum likelihood estimation (MLE) and empirical cumulative density functions tests (ECDF). Simulations shows MLE method to be the better procedure to but not in all analyzed circumstances, moreover the parameter dependency of confidence bounds has been highlighted. A low level way for the detection of warranty bias is introduced.