Monte carlo comparison of estimation methods for additive two-way tables

We considered the problem of estimating effects in the following linear model for data arranged in a two-way table: Response = Common effect + Row effect + Column effect + Residual. This work was occasioned by a project to analyse Federal Aviation Administration (FAA) data on daily temporal deviations from flight plans for commercial US flights, with rows and columns representing origin and destination airports, respectively. We conducted a large Monte Carlo study comparing the accuracy of three methods of estimation: classical least squares, median polish and least absolute deviations (LAD). The experiments included a wide spectrum of tables of different sizes and shapes, with different levels of non-linearity, noise variance, and percentages of empty cells and outliers. We based our comparison on the accuracy of the estimates and on computational speed. We identified factors that significantly affect accuracy and speed, and compared the methods based on their sensitivity to these factors. We concluded that there is no dominant method of estimation and identified conditions under which each method is most attractive.     

社科论文写作中对图表的几点要求

在科技论文写作中,图表因其具有示意性、形象性等优点而被广泛采用.在应用图表过程中,作者常常会因对图表的逻辑规范了解较少而出现各种各样的问题,对图表的规范进行分析和研究具有重要的现实意义.     

Simplifying the calculation of the P-value for Barnard's test and its derivatives

Unconditional non-asymptotic methods for comparing two independent binomial proportions have the drawback that they take a rather long time to compute. This problem is especially acute in the most powerful version of the method (Barnard, 1947). Thus, despite being the version which originated the method, it has hardly ever been used. This paper presents various properties which allow the computation time to be drastically reduced, thus enabling one to use not only the more traditional and simple versions given by McDonald et al. (1977) and Garside and Mack (1967), but also the more complex original version of Barnard (1947).     

Comparing clinical data with administrative data for producing acute myocardial infarction report cards

Summary.  We compared measures of hospital performance by using both administrative and clinical data sources. Hospital-specific mortality outcomes on 10086 patients who had been admitted to 102 hospitals with a diagnosis of acute myocardial infarction in Ontario, Canada, were used as a test-case. Four and six hospitals were identified as having mortality that was statistically significantly higher than expected by using administrative and clinical data respectively, when model-based indirect standardization was used. When using random-effects models, zero and two hospitals were identified as having significantly higher mortality by using administrative and clinical data respectively. Approximately one in four hospitals changed at least two decile rankings when clinical data were used compared with when administrative data were used.     

Iterative Scaling in Curved Exponential Families

The paper describes a generalized iterative proportional fitting procedure that can be used for maximum likelihood estimation in a special class of the general log‐linear model. The models in this class, called relational, apply to multivariate discrete sample spaces that do not necessarily have a Cartesian product structure and may not contain an overall effect. When applied to the cell probabilities, the models without the overall effect are curved exponential families and the values of the sufficient statistics are reproduced by the MLE only up to a constant of proportionality. The paper shows that Iterative Proportional Fitting, Generalized Iterative Scaling, and Improved Iterative Scaling fail to work for such models. The algorithm proposed here is based on iterated Bregman projections. As a by‐product, estimates of the multiplicative parameters are also obtained. An implementation of the algorithm is available as an R‐package.     

Mixed estimation of old‐age mortality

The estimation of the mortality of the “oldest old”; is subject to considerable random error, but important prior information exists that can be used to make the estimates more robust. Mixed estimation is a method of incorporating auxiliary information into the statistical estimation of linear models. We extend the method to cover general maximum likelihood estimation, and show that the mixed estimator can be represented approximately as a weighted average of the purely data based estimator and the auxiliary estimator. The methods can be applied to the analysis of the old‐age mortality via logistic and Poisson regression. A major advantage of the mixed estimator is the simplicity with which it can incorporate partial prior information. Moreover, no special software is needed in the fitting. We show how the targeting methods of Coale and Kisker can be represented as mixed estimation in a natural way that is more flexible than the original proposal. We also derive empirical estimates of the target information based on pooled data from several countries with high quality data. We consider the mortality of Finland at ages 80 +, study the reliability of the evidence of mortality crossover, and derive estimates of life expectancy at age 100.     

On the exact distribution of the likelihood ratio criterion for testing

The exact null distribution of the likelihood ratio criter- 2 ion for testing the hypothesis H: y = y~; z = a I, a unknown and UQ a given known vector against the alternative A =f H in a p-vari- ate normal population N (y,z) has been derived in the form of Meijer's G-function using mellin integral transform and also in a chisquare series form. Asymptotic behavior of the distribution of -2 log L has also been discussed. Percentage points for p=2(l)10for various level of significance and various degrees of freedom have been computed, but only selected tables have been presented in this paper.     

Tables to facilitate fitting lu distributions

This paper is concerned with estimating the parameters of Tadikamalla-Johnson's L_Udistributions based on the method of moments. Tables of the parameters of the L_U distribution are given for selected values of skewness (0.0(0.05) 1.0(0.1)2.0) and for twenty values of kurtosis at intervals of 0.2. The construction and use of these tables is explained with a numerical example.     

Joint-regression Analysis for Incomplete Two-way Tables

Joint-regression analysis of incomplete two-way tables of genotype-by-environment data is usually carried out using the iterative method of Digby (1979). However, the nonlinear analysis method of Ng & Grunwald (1997) can also be used. The speed of convergence for the Ng & Grunwald method compares favourably with that of Digby, and more importantly the method gives valid inference and hence should be preferred.     

混合型P2P网络中的基于Gossip的动态自适应算法

为了提高搜索的性能,出现了混合P2P网络搜索方法。在混合P2P网络中,关键的问题在于确定资源的流行程度。针对该问题,该文提出了一种基于Gossip的动态自适应算法(DAHG),通过抛硬币操作估计节点加入和离开P2P网络时所带入和带走的文档副本数;并通过Gossips传递估计值,获得资源的流行程度。仿真结果表明,该算法具有良好的适应性,能够反映P2P网络中节点和资源的动态性,从而选择出正确的资源搜索方法,减少搜索响应时间和提高资源的命中率。