Comparing Performance of Multiclass Classification Systems with ROC Manifolds: When Volume and Correct Classification Fails

Higher dimensional surfaces are used to examine the diagnostic performance of multiclass classification systems. These surfaces are extensions of the ROC curve and are known as ROC surfaces or manifolds. Manifolds may be constructed from either the correct classifications or from the misclassifications of the diagnostic system. Comparisons of the usefulness of each of these ROC manifolds with respect to the performance of the diagnostic system are made with emphasis on inferences from volume under the surface and optimal operating points (thresholds) of the system. Recommendations for when to use each type of ROC manifold and performance measure are discussed.     

Reliability and effects of muscular pretension on isometric strength of older adults

There were two objectives to this study: (a) to determine the reliability of measuring isometric maximum torque (MT) and maximum rate of torque development (MRTD) of the knee extensor and flexor muscles and (b) to assess the effects of different muscular pretensions on MT and MRTD of older adults. Maximum strength of 35 untrained healthy older adults (aged 66.9?±?6.6 years; height 1.72?±?0.07 m; weight 75.2?±?9.6 kg) was measured. Strength measurement was performed with the Dr. Wolff ISO Check using a piezo-electric force transducer. MT intra-day coefficient of variation (CV) was between 3.30 and 12.40 % for both measurement sessions. Intraclass correlation coefficients (ICC) ranging from 0.71 to 0.99 showed good relative reliability. MT test–retest reliability (between sessions 1 and 2) of knee extension reached a CV between 7.10 and 9.30 % and high ICC values. The measurement of the knee flexion achieved an acceptable CV (9.40 %) at 40 % muscular pretension. MRTD was negatively influenced by muscular pretension (p <0.05). In conclusion, MT could be measured reliably over different trials using isometric strength measurement. The test–retest reliability was acceptable at knee extension with 10 to 40 % muscular pretension and with 40 % muscular pretension at knee flexion. A higher muscular pretension has no impact on MT but decreases MRTD in older adults.     

Zur Messung von Forschungsqualit?t

This paper deals with the Forschungsrating (RWR) by the German Wissenschaftsrat (WR) in sociology. The WR??s objective in conducting this pilot study was to test a practice which allows a sophisticated assessment of the research facilities in Germany. Basically, two theses are exhibited towards the RWR in this paper: First, it seems possible that the published results are neither objective, nor reliable or valid because of implicit assessment procedures by the reviewers. Second we ask, whether a simple model of scientific quality, composed of a few quantitative indicators, yields similar results compared with the RWR. To empirically address these aspects, data of the RWR and our own standardized survey on the reputation of German research facilities is analyzed. Furthermore we conducted interviews with members of the group of reviewers. The results show that problems concerning the objectivity of analysis cannot be excluded. Our general conclusion about the reliability of the measurement, limited to the inner consistency of the quantitative indicators, is positive. A satisfactory reproduction of the RWR??s results solely based on quantitative data cannot be achieved. This is interpreted as a sign of minor validity. It is concluded that the amount and quality of data collected in the RWR, as well as the standardization of the rating process need to be improved.     

On the application of new tests for structural changes on global minimum-variance portfolios

We investigate if portfolios can be improved if the classical Markowitz mean–variance portfolio theory is combined with recently proposed change point tests for dependence measures. Taking into account that the dependence structure of financial assets typically cannot be assumed to be constant over longer periods of time, we estimate the covariance matrix of the assets, which is used to construct global minimum-variance portfolios, by respecting potential change points. It is seen that a recently proposed test for changes in the whole covariance matrix is indeed partially useful whereas pairwise tests for variances and correlations are not suitable for these applications without further adjustments.     

Hierarchie und Entlohnung

In this paper, we suggest a wage scheme that accounts for the hierarchical structure of an enterprise. We employ concepts of cooperative game theory and modify the van den Brink (2008) approach. Besides results on how the hierarchy affects wage differentials between levels of the hierarchy, we deal with the allocation of employees to the different levels.     

Attraktivität, Reziprozität und Lehrveranstaltungsevaluation

In this article we investigate the relationship between academic teachers’ attractiveness and their students’ evaluations of teaching. Furthermore it is examined, whether teachers’ grading influences students’ evaluations of teaching. Using data from an experimental design, hypotheses about attractiveness effects, frog-pond-effects, moderating gender effects and influences of grading are empirically tested. Controlling for confounding factors we find a very weak attractiveness effect, which can partly, but not only be ascribed to beauty’s productivity enhancing influence. Thereby, independently from the raters’ sex, female teachers profit from their looks. In contrast to that attractive male teachers are even sanctioned more strongly by their students for difficult tests than unattractive teachers. The frog-pond-hypothesis cannot empirically be supported: the attractiveness of teachers’ colleagues has no influence on the attractiveness effect. However, reciprocal evaluation behavior can consistently be found in the theoretical expected manner. It is substantially more relevant for students’ evaluations of teaching than attractiveness effects are. Alternative explanations for this grading effect can be excluded.     

Compliance check of gambler and youth protection in German amusement arcades: a pilot study

German gambling operators are required to identify problem gamblers at an early stage and exclude them from gambling on their premises. The main aim of this pilot study was to investigate the compliance of staff in Bremen’s amusement arcades. While visiting 29 of 111 amusement arcades, covert gamblers simulated signs of problematic gambling behaviour, arranged self-exclusions, monitored observance of the exclusion list and made fieldnotes about the staff’s reactions. The application of age checks was also tested. The findings showed that only in six cases (5%) did the staff adequately respond to evident signs of problematic gambling behaviour. Although 18 of 29 cases (62%) eventually resulted in self-exclusions or ban-like agreements, in some instances it was necessary to overcome major obstacles that emanated from the staff’s behaviour. During 15 subsequent follow-up visits, 13 banned mystery gamblers (87%) had no problems gambling on amusement-with-prizes (AWP) machines. Finally, just 26% of the 20- to 25-year-old mystery gamblers had their ID checked. The results indicate the necessity of regular external evaluations regarding the implementation of gambling harm-minimization measures in order to identify any shortcomings. In addition, the level of staff compliance may be increased by amending current legislation (e.g. by introducing a nationwide self-exclusion programme).     

Migration and Pacific Mortality: Estimating Migration Effects on Pacific Mortality Rates Using Bayesian Models

Pacific people living in New Zealand have higher mortality rates than New Zealand residents of European/Other ethnicity. The aim of this paper is to see whether Pacific mortality rates vary by natality and duration of residence. We used linked census-mortality information for 25- to 74-year-olds in the 2001 census followed for up to three years. Hierarchical Bayesian modeling provided a means of handling sparse data. Posterior mortality rates were directly age-standardized. We found little evidence of mortality differences between the overseas-born and the New Zealand–born for all-cause, cancer, and cardiovascular disease (CVD) mortality. However, we found evidence for lower all-cause (and possibly cancer and CVD) mortality rates for Pacific migrants resident in New Zealand for less than 25 years relative to those resident for more than 25 years. This result may arise from a combination of processes operating over time, including health selection effects from variations in New Zealand’s immigration policy, the location of Pacific migrants within the social, political, and cultural environment of the host community, and health impacts of the host culture. We could not determine the relative importance of these processes, but identifying the (modifiable) drivers of the inferred long-term decline in health of the overseas-born Pacific population relative to more-recent Pacific migrants is important to Pacific communities and from a national health and policy perspective.