چكيده به لاتين
Abstract:
Experiments where the response of interest is a profile are used in most engineering applications. In this experiments, for each setting of design factors, responses are collected over an interval of some continuous index, such as time or temperature.
In longitudinal studies, the access to variations within output profile and the detection of effective factors on response variable is possible through repeated measurments of experimental units over time,while in cross - sectional studies variations between profiles are computable, merely.
Typically, Generalized Linear Models are common methods for longitudinal data analysis, including wide range of response distribution involves exponential distribution family and use maximum likelihood approach for regression coefficients estimation. Generalized Linear Mixed Model is an extention to generalized linear model in which the linear predictor contains random effects in addition to usual fixed effects.
The joint optimization plot, graphically displays the result of optimization for multiple responses. This result of optimization is a setting of design factors, which produces responses on target values.
In this research, Generalized Linear Mixed Model will be used for modelling profile response and hierarchical-likelihood approach, that is more flexible for such studies, will be used for estimation of regression coefficients.
Then, joint optimization plot will be used for finding the optimal setting of control factors that leads to pre-specified target values of responses.
Keywords: Profile response experiments, Robust designs, Generalized Linear Models, Hierarchical- likelihood, Joint Optimization Plot
