Sample size determination typically relies on a power analysis based on a frequentist conditional approach. This latter can be seen as a particular case of the two-priors approach, which allows to build four distinct ...Sample size determination typically relies on a power analysis based on a frequentist conditional approach. This latter can be seen as a particular case of the two-priors approach, which allows to build four distinct power functions to select the optimal sample size. We revise this approach when the focus is on testing a single binomial proportion. We consider exact methods and introduce a conservative criterion to account for the typical non-monotonic behavior of the power functions, when dealing with discrete data. The main purpose of this paper is to present a Shiny App providing a user-friendly, interactive tool to apply these criteria. The app also provides specific tools to elicit the analysis and the design prior distributions, which are the core of the two-priors approach.展开更多
It is quite common in statistical modeling to select a model and make inference as if the model had been known in advance;i.e. ignoring model selection uncertainty. The resulted estimator is called post-model selectio...It is quite common in statistical modeling to select a model and make inference as if the model had been known in advance;i.e. ignoring model selection uncertainty. The resulted estimator is called post-model selection estimator (PMSE) whose properties are hard to derive. Conditioning on data at hand (as it is usually the case), Bayesian model selection is free of this phenomenon. This paper is concerned with the properties of Bayesian estimator obtained after model selection when the frequentist (long run) performances of the resulted Bayesian estimator are of interest. The proposed method, using Bayesian decision theory, is based on the well known Bayesian model averaging (BMA)’s machinery;and outperforms PMSE and BMA. It is shown that if the unconditional model selection probability is equal to model prior, then the proposed approach reduces BMA. The method is illustrated using Bernoulli trials.展开更多
In several instances of statistical practice, it is not uncommon to use the same data for both model selection and inference, without taking account of the variability induced by model selection step. This is usually ...In several instances of statistical practice, it is not uncommon to use the same data for both model selection and inference, without taking account of the variability induced by model selection step. This is usually referred to as post-model selection inference. The shortcomings of such practice are widely recognized, finding a general solution is extremely challenging. We propose a model averaging alternative consisting on taking into account model selection probability and the like-lihood in assigning the weights. The approach is applied to Bernoulli trials and outperforms Akaike weights model averaging and post-model selection estimators.展开更多
In applications, the traditional estimation procedure generally begins with model selection.Once a specific model is selected, subsequent estimation is conducted under the selected model withoutconsideration of the un...In applications, the traditional estimation procedure generally begins with model selection.Once a specific model is selected, subsequent estimation is conducted under the selected model withoutconsideration of the uncertainty from the selection process. This often leads to the underreportingof variability and too optimistic confidence sets. Model averaging estimation is an alternative to thisprocedure, which incorporates model uncertainty into the estimation process. In recent years, therehas been a rising interest in model averaging from the frequentist perspective, and some importantprogresses have been made. In this paper, the theory and methods on frequentist model averagingestimation are surveyed. Some future research topics are also discussed.展开更多
Parameter estimation is always a difficult issue for crop model users, and inaccurate parameter values will result in deceptive model predictions. Parameter values may vary with different inversion methods due to equi...Parameter estimation is always a difficult issue for crop model users, and inaccurate parameter values will result in deceptive model predictions. Parameter values may vary with different inversion methods due to equifinality and differences in the estimating processes. Therefore, it is of great importance to evaluate the factors which may influence parameter estimates and to make a comparison of the current widely-used methods. In this study, three popular frequentist methods(SCE-UA, GA and PEST) and two Bayesian-based methods(GLUE and MCMC-AM) were applied to estimate nine cultivar parameters using the ORYZA(v3) Model. The results showed that there were substantial differences between the parameter estimates derived by the different methods, and they had strong effects on model predictions. The parameter estimates given by the frequentist methods were obviously sensitive to initial values, and the extent of the sensitivity varied with algorithms and objective functions. Among the frequentist methods, the SCE-UA was recommended due to the balance between stable convergence and high efficiency. All the parameter estimates remarkably improved the goodness of model-fit, and the parameter estimates derived from the Bayesian-based methods had relatively worse performance compared to the frequentist methods. In particular, the parameter estimates with the highest probability density of posterior distributions derived from the MCMC-AM method(MCMC_P_(max)) led to results equivalent to those derived from the frequentist methods, and even better in some situations. Additionally, model accuracy was greatly influenced by the values of phenology parameters in validation.展开更多
Bayesian model averaging (BMA) is a popular and powerful statistical method of taking account of uncertainty about model form or assumption. Usually the long run (frequentist) performances of the resulted estimator ar...Bayesian model averaging (BMA) is a popular and powerful statistical method of taking account of uncertainty about model form or assumption. Usually the long run (frequentist) performances of the resulted estimator are hard to derive. This paper proposes a mixture of priors and sampling distributions as a basic of a Bayes estimator. The frequentist properties of the new Bayes estimator are automatically derived from Bayesian decision theory. It is shown that if all competing models have the same parametric form, the new Bayes estimator reduces to BMA estimator. The method is applied to the daily exchange rate Euro to US Dollar.展开更多
文摘Sample size determination typically relies on a power analysis based on a frequentist conditional approach. This latter can be seen as a particular case of the two-priors approach, which allows to build four distinct power functions to select the optimal sample size. We revise this approach when the focus is on testing a single binomial proportion. We consider exact methods and introduce a conservative criterion to account for the typical non-monotonic behavior of the power functions, when dealing with discrete data. The main purpose of this paper is to present a Shiny App providing a user-friendly, interactive tool to apply these criteria. The app also provides specific tools to elicit the analysis and the design prior distributions, which are the core of the two-priors approach.
文摘It is quite common in statistical modeling to select a model and make inference as if the model had been known in advance;i.e. ignoring model selection uncertainty. The resulted estimator is called post-model selection estimator (PMSE) whose properties are hard to derive. Conditioning on data at hand (as it is usually the case), Bayesian model selection is free of this phenomenon. This paper is concerned with the properties of Bayesian estimator obtained after model selection when the frequentist (long run) performances of the resulted Bayesian estimator are of interest. The proposed method, using Bayesian decision theory, is based on the well known Bayesian model averaging (BMA)’s machinery;and outperforms PMSE and BMA. It is shown that if the unconditional model selection probability is equal to model prior, then the proposed approach reduces BMA. The method is illustrated using Bernoulli trials.
文摘In several instances of statistical practice, it is not uncommon to use the same data for both model selection and inference, without taking account of the variability induced by model selection step. This is usually referred to as post-model selection inference. The shortcomings of such practice are widely recognized, finding a general solution is extremely challenging. We propose a model averaging alternative consisting on taking into account model selection probability and the like-lihood in assigning the weights. The approach is applied to Bernoulli trials and outperforms Akaike weights model averaging and post-model selection estimators.
基金supported by the National Natural Science Foundation of China under Grant Nos. 70625004, 10721101, and 70221001
文摘In applications, the traditional estimation procedure generally begins with model selection.Once a specific model is selected, subsequent estimation is conducted under the selected model withoutconsideration of the uncertainty from the selection process. This often leads to the underreportingof variability and too optimistic confidence sets. Model averaging estimation is an alternative to thisprocedure, which incorporates model uncertainty into the estimation process. In recent years, therehas been a rising interest in model averaging from the frequentist perspective, and some importantprogresses have been made. In this paper, the theory and methods on frequentist model averagingestimation are surveyed. Some future research topics are also discussed.
基金supported by the National Natural Science Foundation of China(NSFC 51909004)。
文摘Parameter estimation is always a difficult issue for crop model users, and inaccurate parameter values will result in deceptive model predictions. Parameter values may vary with different inversion methods due to equifinality and differences in the estimating processes. Therefore, it is of great importance to evaluate the factors which may influence parameter estimates and to make a comparison of the current widely-used methods. In this study, three popular frequentist methods(SCE-UA, GA and PEST) and two Bayesian-based methods(GLUE and MCMC-AM) were applied to estimate nine cultivar parameters using the ORYZA(v3) Model. The results showed that there were substantial differences between the parameter estimates derived by the different methods, and they had strong effects on model predictions. The parameter estimates given by the frequentist methods were obviously sensitive to initial values, and the extent of the sensitivity varied with algorithms and objective functions. Among the frequentist methods, the SCE-UA was recommended due to the balance between stable convergence and high efficiency. All the parameter estimates remarkably improved the goodness of model-fit, and the parameter estimates derived from the Bayesian-based methods had relatively worse performance compared to the frequentist methods. In particular, the parameter estimates with the highest probability density of posterior distributions derived from the MCMC-AM method(MCMC_P_(max)) led to results equivalent to those derived from the frequentist methods, and even better in some situations. Additionally, model accuracy was greatly influenced by the values of phenology parameters in validation.
文摘Bayesian model averaging (BMA) is a popular and powerful statistical method of taking account of uncertainty about model form or assumption. Usually the long run (frequentist) performances of the resulted estimator are hard to derive. This paper proposes a mixture of priors and sampling distributions as a basic of a Bayes estimator. The frequentist properties of the new Bayes estimator are automatically derived from Bayesian decision theory. It is shown that if all competing models have the same parametric form, the new Bayes estimator reduces to BMA estimator. The method is applied to the daily exchange rate Euro to US Dollar.
