Tore Schweder

Health complaints and job stress in Norwegian physicians: the use of an overlapping questionnaire design

An extensive research program has been undertaken in Norway on physician health, sickness, working conditions and quality of life. Data are collected from cross-sectional and longitudinal prospective and retrospective surveys, qualitative studies, and vital statistics. This paper presents findings on subjectively experienced health problems, emotional distress, experienced job stress and job satisfaction, based on an extensive cross-sectional postal questionnaire study in 1993. An overlapping questionnaire design was used to allow many relationships to be estimated without exhausting the recipients. 9266 active physicians were included, which comprises close to the total Norwegian physician work-force minus a representative sample of 2100, used for other studies. The primary questionnaire was returned by 6652 (71.8%), the great majority of which also returned three secondary questionnaires. The results indicate that health complaints were significantly more frequent in female physicians and decreased with age. Low job satisfaction, high job stress, and emotional distress were all found to be significant predictors of subjective health complaints, as measured by the Ursin Health Inventory.

Confidence and Likelihood

Confidence intervals for a single parameter are spanned by quantiles of a confidence distribution, and one-sided p-values are cumulative confidences. Confidence distributions are thus a unifying format for representing frequentist inference for a single parameter. The confidence distribution, which depends on data, is exact (unbiased) when its cumulative distribution function evaluated at the true parameter is uniformly distributed over the unit interval. A new version of the Neyman-Pearson lemma is given, showing that the confidence distribution based on the natural statistic in exponential models with continuous data is less dispersed than all other confidence distributions, regardless of how dispersion is measured. Approximations are necessary for discrete data, and also in many models with nuisance parameters. Approximate pivots might then be useful. A pivot based on a scalar statistic determines a likelihood in the parameter of interest along with a confidence distribution. This proper likelihood is reduced of all nuisance parameters, and is appropriate for meta-analysis and updating of information. The reduced likelihood is generally different from the confidence density. Confidence distributions and reduced likelihoods are rooted in Fisher-Neyman statistics. This frequentist methodology has many of the Bayesian attractions, and the two approaches are briefly compared. Concepts, methods and techniques of this brand of Fisher-Neyman statistics are presented. Asymptotics and bootstrapping are used to find pivots and their distributions, and hence reduced likelihoods and confidence distributions. A simple form of inverting bootstrap distributions to approximate pivots of the abc type is proposed. Our material is illustrated in a number of examples and in an application to multiple capture data for bowhead whales.

Some “Optimal” Methods to Detect Structural Shift or Outliers in Regression

Abstract A Neyman-Pearson approach is taken to the problem of detecting structural shifts in naturally ordered regression problems. When the variance is known, backwards CUSUM methods are shown to maximize average power, and their application is discussed. Two methods with optimality properties for outlier detection are developed, assuming that the observations may be divided into two parts, where the first part satisfies the model assumptions, while outliers may be present in the other.

Suicide rates from 1960 to 1989 in Norwegian physicians compared with other educational groups

The aim of the present study is to compare suicide rates between 1960 and 1989 for Norwegian physicians with corresponding rates for other Norwegians with and without university education, by age, gender, and five-year period, based on death certificates for all Norwegians who died in the period 1960-1989. There were 82 registered physician suicides, of which 9 were female, 265 suicides by persons with other university education, and 11,165 by persons with no university education. Suicide rate is measured in number of deaths per 100,000 person years. Crude suicide rates were 47.7 (95% CI 37.7-60.4) for male physicians, 20.1 (17.7-22.9) for other male university graduates, and 22.7 (22.2-23.2) for men with no university education. The corresponding figures for females were 32.3 (15.8-63.7), 13.0 (8.4-19.8) and 7.7 (7.5-8.0). Both for males and females, suicide rates, controlled for age and period, were significantly higher for physicians than for persons with other or no university education. Poisson modelling showed that the risk of suicide for male physicians has the same age pattern as for other males with higher education. In 1985-89 the suicide rate for male physicians increased nearly linearly from about 35 at the age 35-40 to about 100 at the age 75-79, which was almost three times higher than for the other male university graduates. For the age group 50-54 the estimated rate increases from about 50 in 1960-64 to about 90 in 1985-89. For the female physicians, the low number of cases prevents reliable estimation of trends. For male physicians, the trend from 1960 to 1989 is increasing. The estimated risk for a single physician to commit suicide was almost 5 times that of a married or co-habitant colleague. For 52% of the male and 85% of the female physicians the suicide method was poisoning. This is about twice the rates in the general population.

A Simple Test for a Set of Sums of Squares

An alternative to Bartletts test is presented, with a ranking method for separating the larger sums of squares from a homogeneous null group. When the test shows heterogeneity one sum of squares is excluded and the test applied again to the remaining set.

Likelihood-based inference for clustered line transect data

The uncertainty in estimation of spatial animal density from line transect surveys depends on the degree of spatial clustering in the animal population. To quantify the clustering we model line transect data as independent thinnings of spatial shot-noise Cox processes. Likelihood-based inference is implemented using Markov chain Monte Carlo (MCMC) methods to obtain efficient estimates of spatial clustering parameters. Uncertainty is addressed using parametric bootstrap or by consideration of posterior distributions in a Bayesian setting. Maximum likelihood estimation and Bayesian inference are compared in an example concerning minke whales in the northeast Atlantic.

Phenotypic evolution studied by layered stochastic differential equations

Time series of cell size evolution in unicellular marine algae (division Haptophyta; Coccolithus lineage), covering 57 million years, are studied by a system of linear stochastic differential equations of hierarchical structure.The data consists of size measurements of fossilized calcite platelets (coccoliths) that cover the living cell, found in deep-sea sediment cores from six sites in the world oceans and dated to irregularly points in time. To accommodate biological theory of populations tracking their fitness optima, and to allow potentially interpretable correlations in time and space, the model framework allows for an upper layer of partially observed site-specific population means, a layer of site-specific theoretical fitness optima and a bottom layer representing environmental and ecological processes. While the modeled process has many components, it is Gaussian and analytically tractable. A total of 710 model specifications within this framework are compared and inference is drawn with respect to model structure, evolutionary speed and the effect of global temperature.

Fisherian or Bayesian methods of integrating diverse statistical information

Directly observed data are in many cases combined with diverse indirect information to draw inference on parameters of interest to the fishery scientist. The indirect information might be based on previous data, analogous data or the researchers expert judgement. The Bayesian prior distribution is the most common concept for representing such indirect information, and the Bayesian paradigm is gaining popularity. An alternative methodology based on the likelihood principle is presented and compared to the Bayesian. In the tradition of R.A. Fisher, the method concentrates on the likelihood function, without bringing in prior distributions that are not based on data. To provide for the integration of relevant indirect statistical information into the likelihood function, the concept of indirect likelihood is proposed. The indirect likelihood is treated as an ordinary independent component of the likelihood. If the indirect likelihood of a parameter is based on previous data, the inclusion of the indirect likelihood in the new study amounts to combining the old and the new data. The two methods are explained and compared, and it is argued that the likelihood method often is advantageous in the scientific context.

In utero and lactational exposure to PCB 118 and PCB 153 alter ovarian follicular dynamics and GnRH-induced luteinizing hormone secretion in female lambs

The effects of in utero and lactational exposure to two structurally different polychlorinated biphenyl (PCB) congeners on follicular dynamics and the pituitary‐gonadal axis in female lambs were investigated. Pregnant ewes received corn oil, PCB 118, or PCB 153, and offspring was maintained until 60 days postpartum. Ovarian follicles were quantified using stereology. Plasma luteinizing hormone (LH) and follicle‐stimulating hormone (FSH) were measured using radioimmunoassay before and after administration of a gonadotropin releasing hormone (GnRH) analog. PCB 118 exposure increased numbers of transitional, secondary, and the sum of secondary, early antral, and antral (Σsecondary‐antral) follicles, PCB 153 exposure only increased the number of primary follicles. GnRH‐induced LH levels were significantly elevated in the PCB 153 exposure group. We conclude that PCB 153 and PCB 118 alter follicular dynamics in lambs and modulate the responsiveness of the pituitary gland to GnRH.

Inference about the Ratio of Two Parameters, with Application to Whale Censusing

Abstract Inference for the quotient of two parameters estimated separately may be obtained by the delta method. When the distribution of linear transformations involving the numerator and the denominator is available, more exact and elementary methods may be used. Non-Bayesian and Bayesian approaches are developed. The application of the methods to estimating the stock abundance of Northeastern Atlantic minke whales, where the ratio is a raw estimate divided by a measure of observation efficiency, is explained and discussed. The Bayesian approach allows exact inference in quite general situations using only a single, rapidly implemented, one-dimensional numerical integration. A simple analytic approximation is given for the common situation where the joint posterior distribution of the numerator and denominator can be approximated by a normal distribution that gives very little probability to negative values of the denominator. The Bayesian approach also permits the incorporation of model uncertainty (or ...