Hinge estimators of location: robust to asymmetry.

Robust estimators have been developed and tested for symmetric distributions via simulation studies. The primary objective of these robust estimators was to show that these estimators had a higher efficiency than the sample mean over these symmetric distributions. Little attention has been given to how these estimators perform on data that are from asymmetric distributions or from distributions that have inherent anomalies-so called 'messy data'. This study is intended to supplement previous studies by examining the behavior of several robust estimators over asymmetric distributions. The objective is to demonstrate several adaptive 'asymmetric' robust estimators which utilize sample selector statistics to identify the underlying distribution and to demonstrate the efficiency of these adaptive estimators. From a methodology point rather than a theoretical basis, reasonable alternatives should be available. In the asymmetric data distributions faced on a daily basis, estimators that adapt themselves to the data may be formulated and used. We recommend the use of the following algorithm in examining data sets: (a) compute the ancillary statistics-skewness and tail-length to classify the data distribution; (b) analyze each data set using at least one alternative estimator to the usual XM; (c) if the results are similar, report the XM analysis; (d) if the results are dissimilar, report the alternative analysis and the reasons for using the alternative analysis (i.e. t-tests based on a T alpha, HQ1, HQ2, or SK5).

Monte Carlo generation of SB and SU distributions.

Generating random variables from specific bounded or unbounded distributions is a problem frequently encountered in Monte Carlo studies. Of particular interest is that of transforming variables from a standard normal distribution to a given Johnson's SB or SU distribution with specified (gamma, delta) parameters. We use a composite standard normal generator N (0.1) to generate Johnson's SB or SU distributions with (gamma, delta) parameters. Goodness-of-fit tests and graphical illustrations demonstrate the adequacy of the empirical distributions.

Evaluation of Monte Carlo generation of long-tailed symmetric and contaminated symmetric distributions.

Generating random variables from a specific distribution, whether symmetric or asymmetric, is a concern of investigators involved in Monte Carlo studies. Of particular interest to those concerned with robustness is the generation of contaminated symmetric distributions such as those used in the Princeton Robustness Study. A reliable composite uniform U(0,1) generator is described and algorithms for transforming U(0,1) to symmetric long-tailed and contaminated symmetric distributions are given. Goodness-of-fit tests and graphical illustrations demonstrate the adequacy of the empirical distributions.

Evaluation of Monte Carlo generation of continuous asymmetric distributions.

Reliable methods for generating pseudo-random numbers from specific distributions are increasingly important in all branches of applied mathematics. In Monte Carlo studies, generating random variables from specific continuous probability distributions, whether symmetric or asymmetric is a fundamental consideration. A composite uniform U(0,1) generator algorithm is described and statistically tested. Algorithms for transforming the U(0,1) to a set of selected continuous probability distributions are also validated.

Robust alternatives to traditional analysis of variance: Welch W*, James J I*, James J II*, Brown-Forsythe BF*.

Contaminated data sets commonly appear in biomedical sciences. Traditional statistical analyses are based on a 'normality' assumption and homogeneity of variances assumption. If either or both assumptions are violated, traditional statistical procedures may give an inflated type I error rate and are then not robust. Robust alternatives to traditional one-way analysis of variance have better power curves and protect against inflated type I error rates.

Hepatic tumors: magnetic resonance and CT appearance.

The magnetic resonance (MR) features of primary and metastatic hepatic tumors were analyzed and compared with CT findings in 28 patients. MR images were obtained on a 0.35-Tesla superconducting magnet using a variety of spin echo and inversion recovery imaging techniques. The normal liver appeared homogeneous and of moderate intensity. Tumors typically appeared as masses of increased intensity on spin echo images and diminished intensity on inversion recovery images. Tumors had prolonged T1 and T2 times, which varied in different types of tumors and within regions of a single tumor. A specific T1 and/or T2 time could not be ascribed to any tumor. Tumor intensity varied greatly depending on the imaging techniques employed, becoming isointense with normal liver on some imaging sequences. MR and CT detected lesions equally well, but internal architecture and the relationship of tumors to hepatic vascular structures were better displayed on MR.

Nonsurgical approach to congenital scalp and skull defects.

The management of congenital scalp and skull defects, as generally advocated, is surgical. The authors report such a case that was treated conservatively. At her 3-year follow-up review, the patient's scalp and skull defects and other associated cutaneous defects were fully reconstituted. Such a nonoperative approach, while rarely reported, emphasizes the natural course that some of these lesions may follow. The literature on aplasia cutis congenita is briefly reviewed.