On the Choice of Functional Forms: Summary of a Monte Carlo Experiment

The use of flexible functional forms is a standard practice in applied econometrics. Many flexible forms have been proposed. In this study, we investigate the behavior of three of them—the translog, the symmetric McFadden, and the symmetric generalized Barnett. Based on Monte Carlo experiments, we assess the ability of these forms to test theoretical properties and to measure technological characteristics.     

Covariate classification by post–stratification

The plug–in Anderson's covariate classification statistic is constructed on the basis of an initially unclassified training sample by means of posty–stratification. The asymptotic efficiency relative to the discriminant based on an initially classified training sample is evaluated for the case where a covariate is present. Effect of post–stratification is examined.     

Equivariant prediction in finite populations

A theory of equivariant prediction is developed for predicting the population total in finite populations. Minimum risk equivariant predictors (MREP) are derived under the location, scale and locationscale superpopulation models. Under the general linear model, it is shown that the best(linear) unbiased predictor (B(L)UP) is an MREP.     

Discrimination between the complementary log-log and logistic model for ordinal data

An assessment is made of the number of observations on ordinal data necessary for reasonable power in a significance test of the null hypothesis of a logit model versus an alternative of a complementary log-log and vice versa. The conclusion is that quite modest numbers of observations, e.g. 50-100, are adequate provided that the boundaries of the ordinal scale are suitably defined.     

The power functions of the likelihood ratio tests for a simply ordered trend in normal means

Likelihood ratio tests for the homogeneity of k normal means with the alternative restricted by an increasing trend are considered as well as the likelihood ratio tests of the null hypothesis that the means satisfy the trend. While the work is primarily a survey of results concerning the power functions of these tests, the extensions of some results to the case of not necessarily equal sample sizes are presented. For the case of known or unknown population variances, exact expressions are given for the power functions for k=3,4, and approximations are discussed for larger k. The topics of consistency, bias and monotonicity of the power functions are included. Also, Bartholomew's conjectures concerning minimal and maximal powers are investigated, with results of a new numerical study given.     

Estimation of gene/haplotype frequencies in genetic marker systems based on phenotype data

Two equivalent methods (gene counting and maximum likelihood) for estimating gene frequencies in a general genetic marker system based on observed phenotype data are derived. Under the maximum likelihood approach, an expression is given for the estimated covariance matrix from which estimated standard errors of the estimators can be found. In addition, consideration is given to the problem of estimating gene frequencies when there are available several independent population data sets.     

Application of extreme value theory to the bec family of distributions

Arnold and Strauss (1988) derived a family of bivariate life distributions having the property that the conditional distributions are exponential. Asymptotic distributions for the marginal and bivariate extremes for this family of distributions are derived employing the asymptotic theory of extreme order statistics.     

Applications of schur theory to reliability inference for k of n systems

The results of Hoeffding (1956), Pledger and Proschan (1971), Gleser (1975) and Boland and Proschan (1983) are used to obtain Buehler (1957) 1-α lower confidence limits for the reliability of k of n systems of independent components when the subsystem data have equal sample sizes and the observed failures satisfy certain conditions. To the best of our knowledge, for k ≠ 1 or n, this is the first time the exact optimal lower confidence limits for system reliability have been given. The observed failure vectors are a generalization of key test results for k of n systems, k ≠ n (Soms (1984) and Winterbottom (1974)). Two examples applying the above theory are also given.     

Some further aspects of the löwner-ordering antitonicity of the moore-penrose inverse

A survey is given of known proofs of the antitonicity of the inverse matrix function for positive definite matrices w.r.t. the Lowner partial ordering, and of the corresponding result for the Moore-Penrose inverse of nonnegative definite matrices [the theorem of Milliken and Akdeniz (1977)]. A short new proof of the latter result is obtained by employing an extremal representation of a nonnegative definite quadratic form. Another proof of this result involving Schur complements is also given, and is seen to be extendable to the case of symmetric (not necessarily nonnegative definite) matrices. A geometrical interpretation of Milliken and Akdeniz's theorem is presented. As an application, the relationship between the concepts of greater (maximum) concentration and smaller (minimum) dispersion is considered for a pair (class) of vector-valued statistics with possibly degenerate distributions.     

Taguchi methods: linear graphs of high resolution

An important reason behind the success of the Taguchi methodology in qual- ity assurance has been the use of statistical methods, presented in a way that is accessible to the nonexpert user. Among the tools used to simplify the sta- tistical design of experiments has been the linear graph, apparently introduced by Taguchi. However, he did not consider the resolution of the corresponding designs (the higher the resolution, the more accurate the conclusions). For example, it will be shown that half of the linear graphs given by Taguchi for the L₁₆(2¹⁵) orthogonal array correspond to designs of resolution III, when designs of resolution IV are available (with the same lines in the linear graphs but with different assignments to the columns of the orthogonal array). A nontraditional but very straightforward method is presented for obtaining the alias chains and the linear graphs corresponding to an orthogonal array. The procedure can be easily understood and employed by nonstatisticians to find an experimental design of the highest possible resolution. The design can be used to obtain products or processes that are robust to variation.