今天介绍的模型平均方法,不同于模型选择依赖于单一模型,而是多个模型的加权组合,不轻易排除任何模型,因此有效降低了有用信息的丢失。这使得估计更加稳健,且保证了较高的预测性能。同时,模型平均给予更好的模型更高的权重,提供了一种保障机制,有效避免了模型选择方法可能存在的缺陷。随着计算机技术的快速发展,模型平均方法作为一种更为复杂的数据挖掘方法将被更多的运用于实际问题中。模型平均方法分为两个方向:贝叶斯模型平均(Bayesian Model Averaging)(BMA) 和频率模型平均(Frequentist Model Averaging)(FMA),下面分别对它们进行介绍。
其中,为参数的有效个数。准则(5)是模型选择中Mallows'准则的推广,其取决于未知的。这个准则是模型选择方法中Mallows 准则的推广,如果我们用估计值代替,那么Mallows准则可用于选择权重向量。通过极小化Mallows准则得到的权重向量为其也被称为Mallows经验权重选择向量,对应的模型平均估计称为Mallows Model Average(MMA)估计。这样,我们可以证明MMA估计在实现最小均方误差(MSE)意义上的渐近最优性,证明过程详见Hansen(2007)[6]。
1.David Madigan and Adrian E Raftery. “Model selection and accounting for model uncertainty in graphical models using Occam’s window”. In: Journal of the American Statistical Association 89.428 (1994), pp. 1535–154
2.David Madigan, Jeremy York, and Denis Allard. “Bayesian Graphical Models for Discrete Data”. In: International Statistical Review / Revue Internationale de Statistique 63.2 (Aug. 1995), p. 215. issn: 03067734. JSTOR: 1403615.
3.Luke Tierney and Joseph B. Kadane. “Accurate Approximations for Posterior Moments and Marginal Densities”. In: Journal of the American Statistical Association 81.393 (Mar. 1986), pp. 82–86. issn: 0162-1459, 1537-274X.
4.Ross H. Taplin. “Robust Likelihood Calculation for Time Series”. In: Journal of the Royal Statistical Society. Series B(Methodological) 55.4 (1993), pp. 829–836.
5.S. T. Buckland, K. P. Burnham, and N. H. Augustin. “Model Selection: An Integral Part of Inference”. In: Biometrics
6.Bruce E. Hansen. “Least Squares Model Averaging”. In: Econometrica 75.4 (July 2007), pp. 1175–1189. issn: 0012-9682, 1468-0262
7.Donald W. K. Andrews. “Heteroskedasticity and Autocorrelation Consistent Covariance Matrix Estimation”. In: Econo- metrica 59.3 (1991), pp. 817–858. issn: 00129682, 14680262
8.Alan T.K. Wan, Xinyu Zhang, and Guohua Zou. “Least squares model averaging by Mallows criterion”. In: Journal of Econometrics 156.2 (June 2010), pp. 277–283. issn: 03044076
9.Bruce E. Hansen. “Least-Squares Forecast Averaging”. In: Journal of Econometrics 146.2 (Oct. 2008), pp. 342–350.
10.Buckland S T, Burnham K P, Augustin N H. Model selection: an integral part of inference[J]. Biometrics, 1997: 603-618.
11.Jennifer A. Hoeting et al. “Bayesian Model Averaging: A Tutorial”. In: Statistical Science 14.4 (1999), pp. 382–401. JSTOR: 2676803.
12.Juditsky A, Nemirovski A. Functional aggregation for nonparametric regression[J]. The Annals of Statistics, 2000, 28(3): 681-712.
13.Homer C, Huang C, Yang L, et al. Development of a 2001 national land-cover database for the United States[J]. 2004.
14.Goldenshluger A. A universal procedure for aggregating estimators[J]. The Annals of Statistics, 2009: 542-568.
15.David Madigan, Jeremy York, and Denis Allard. “Bayesian Graphical Models for Discrete Data”. In: International Statistical Review / Revue Internationale de Statistique 63.2 (Aug. 1995), p. 215. issn: 03067734. JSTOR: 1403615.
16.Enrique Moral-Benito. “MODEL AVERAGING IN ECONOMICS: AN OVERVIEW: MODEL AVERAGING IN ECONOMICS”. In: Journal of Economic Surveys 29.1 (Feb. 2015), pp. 46–75. issn: 09500804.
17.Alan T.K. Wan, Xinyu Zhang, and Guohua Zou. “Least squares model averaging by Mallows criterion”. In: Journal of Econometrics 156.2 (June 2010), pp. 277–283. issn: 03044076.