Explicit Statistical Model Research Articles

Mapping Mediterranean Altimeter Data with a Multiresolution Optimal Interpolation Algorithm

A multiresolution optimal interpolation scheme is described and used to map the sea level anomaly of the Mediterranean Sea based on TOPEX/Poseidon and ERS-1 data. The principal advantages of the multiresolution scheme are its high computational efficiency, the requirement for explicit statistical models for the oceanographic signal and the measurement errors, and the production of error variances for all estimates at multiple scales. A set of MATLAB-callable routines that implement the multiresolution scheme have been made available via anonymous FTP. The oceanographic signal is here modeled as a stationary 1/km process, where k is the horizontal wavenumber. Measurement noise is modeled as the sum of two separate random processes: a Gaussian white noise process and a correlated process of a low wavenumber representing the uncertainties in the orbital position of the satellite and in the atmospheric load corrections. The efficiency of the multiresolution scheme allowed the testing of more than 16 000 sets of hypothesized statistical prior model parameters to determine the most likely parameters. Mapping results with and without low wavenumber error corrections are presented and compared.

Estimation of the coefficient of variation from laboratory analysis of split specimens for quality control in clinical trials

An explicit statistical model is proposed for the coefficient of variation for laboratory analyses of constituents of blood, serum, saliva, or other specimens. A method for computing the maximum likelihood estimate of the key parameter is described, and compared with two simpler noniterative estimates. Validity of the model is explored by analysis of data from the central laboratory of a large cooperative clinical trial. Simulation studies are employed to compare the accuracy of the three estimators of the coefficient of variation. For most laboratory measurements for which the model is valid, one of the two noniterative estimates is nearly as accurate and unbiased as the maximum likelihood estimate.

Optimum beamforming for coherent signal and interferences

An approach is described to optimum beamforming in the presence of correlated interferences completely overcoming the signal cancellation phenomenon. In contrast with classical adaptive beamforming where no assumptions are made on the statistical properties of the interference and noise, other than their being uncorrelated with the desired signal, the present approach is based on an explicit statistical model of the signal environment. The method is adaptive, however, in that the characterization of the signal environment is extracted from the measured array data.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

A statistical model for integrated-circuit yield with clustered flaws

An explicit statistical model for die yield in integrated circuits (ICs) is developed. The model includes the effect of defect clustering, and is based on empirical observations of IC yield patterns. Using the simplest assumptions it leads to the well-known gamma-function yield formula.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Statistical analysis of empirical models fitted by optimization

Practical procedures for statistical inference about the parameters of empirical models are presented. These avoid the necessity of formulating an explicit statistical model and can allow for general statistical dependency within the data used for model fitting. The theory is an asymptotic one, and one which allows the construction of significance tests and confidence regions for subsets of the model parameters. Other results concern the comparison of structurally different models, and of different objective functions for parameter fitting.

Distribution Theory for Glass's Estimator of Effect size and Related Estimators

Glass's estimator of effect size, the sample mean difference divided by the sample standard deviation, is studied in the context of an explicit statistical model. The exact distribution of Glass's estimator is obtained and the estimator is shown to have a small sample bias. The minimum variance unbiased estimator is obtained and shown to have uniformly smaller variance than Glass's (biased) estimator. Measurement error is shown to attenuate estimates of effect size and a correction is given. The effects of measurement invalidity are discussed. Expressions for weights that yield the most precise weighted estimate of effect size are also derived.