Mis-specification Analysis of Generalized Gamma distribution for ALT censored data
Dissertation abstract
My research focuses on mis- specified distribution apply to reliability modeling.
Based on lifetime distributions, log normal and Weibul are one of the most widely used distribution in reliability engineering, location scale models which involves one or more shape parameters also are sometimes useful. Beside additional flexibility for fitting data, such models can provide comparisons of
Weibull and Lognormal models and can be used to examine the robustness of conclusions to plausible variation in the model.
The generalized gamma distribution with three parameters, which include one more “shape”, is one of these models, and includes the Weibul and Lognormal as special cases. This model is originally introduced by specifying with one parameter gamma distribution with index parameter k>0.
However the mean and the variance of the gamma distribution both equal to k, and as “k” increase the gamma does not have limits, so the cases k=1 and k = ∞ give the Weibull and Lognormal distribution, the two parameter gamma distribution also arises as another special cases.
For the model mis - specification when using maximum likelihood techniques for estimation and inference, the quasi-maximum likelihood estimator (QMLE) converges to a well-defined limit.
When the distribution mis– specified, the lifetime quantile often are interested to the practitioner, which predict may differ significantly between wrong distributions for several k.
Accelerate lifetime (ALT) or log – location scale model, denote to the models that usually have the same shape but are separated by a distance, such models are especially useful when lifetimes of individuals can vary by orders of magnitude, many engineering models in which failure is accelerated by thermal, voltage or other stresses are of this type.
Under the specific ALT, we investigate the relative bias and relative variability of lifetime quantile due to distribution mis – specification for censoring data when
generalized
My research focuses on mis- specified distribution apply to reliability modeling.
Based on lifetime distributions, log normal and Weibul are one of the most widely used distribution in reliability engineering, location scale models which involves one or more shape parameters also are sometimes useful. Beside additional flexibility for fitting data, such models can provide comparisons of
Weibull and Lognormal models and can be used to examine the robustness of conclusions to plausible variation in the model.
The generalized gamma distribution with three parameters, which include one more “shape”, is one of these models, and includes the Weibul and Lognormal as special cases. This model is originally introduced by specifying with one parameter gamma distribution with index parameter k>0.
However the mean and the variance of the gamma distribution both equal to k, and as “k” increase the gamma does not have limits, so the cases k=1 and k = ∞ give the Weibull and Lognormal distribution, the two parameter gamma distribution also arises as another special cases.
For the model mis - specification when using maximum likelihood techniques for estimation and inference, the quasi-maximum likelihood estimator (QMLE) converges to a well-defined limit.
When the distribution mis– specified, the lifetime quantile often are interested to the practitioner, which predict may differ significantly between wrong distributions for several k.
Accelerate lifetime (ALT) or log – location scale model, denote to the models that usually have the same shape but are separated by a distance, such models are especially useful when lifetimes of individuals can vary by orders of magnitude, many engineering models in which failure is accelerated by thermal, voltage or other stresses are of this type.
Under the specific ALT, we investigate the relative bias and relative variability of lifetime quantile due to distribution mis – specification for censoring data when
generalized