Generalized additive modelling of sample extremes

Summary.  We describe smooth non-stationary generalized additive modelling for sample extremes, in which spline smoothers are incorporated into models for exceedances over high thresholds. Fitting is by maximum penalized likelihood estimation, with uncertainty assessed by using differences of deviances and bootstrap simulation. The approach is illustrated by using data on extreme winter temperatures in the Swiss Alps, analysis of which shows strong influence of the north Atlantic oscillation. Benefits of the new approach are flexible and appropriate modelling of extremes, more realistic assessment of estimation uncertainty and the accommodation of complex dependence patterns.     

Accurate and efficient construction of bootstrap likelihoods

The simplest construction of bootstrap likelihoods involves two levels of bootstrapping, kernel density estimation, and non-parametric curve-smoothing. We describe more accurate and efficient constructions, based on smoothing at the first level of nested bootstraps and saddlepoint approximation to remove second-level bootstrap variation. Detailed illustrations are given.     

Age-related changes in maximal power and maximal heart rate recorded during a ramped test in 114 cyclists age 15-73 years

This study assessed age-related changes in power and heart rate in 114 competitive male cyclists age 15-73 years. Participants completed a maximal Kingcycle ergometer test with maximal ramped minute power (RMPmax, W) recorded as the highest average power during any 60 s and maximal heart rate (HRmax, beats/min) as the highest value during the test. From age 15 to 29 (n = 38) RMPmax increased by 7.2 W/year (r = .53, SE 49 W, p < .05). From age 30 to 73 (n = 78) RMPmax declined by 2.4 W/year (r = - .49, SE 49 W, p < .05). Heart rate decreased across the full age range by 0.66 beats . min( -1 ) . year( -1 ) (r = -.75, SE 9 beats/min, p < .05). Age accounted for only 25% of the variance in RMPmax but 56% in HRmax. RMPmax was shown to peak at age 30, then decline with age, whereas HRmax declined across the full age range.     

Employee absence, age and tenure: a study of nonlinear effects and trivariate models

Multiple regression techniques were used to explore the nature of the relationships between age, tenure and absence in 2417 British local government workers drawn from three work groups. The data were collected from organizational records and included measures of both non-certified absence and medically certified absence. Linear and curvilinear associations among age, tenure and absence were examined. The possible moderating or mediating role of tenure in the relationship between age and absence was subsequently analysed. The data revealed linear relationships between age and absence that were negative for non-certified absence and positive for certified absence. In contrast, curvilinear relationships were found between tenure and absence that were U-shaped for noncertified absence and inverse U-shaped for certified absence. Tenure was found to moderate but not to mediate the relationship between age and absence. The implications of the results are discussed in the context of the changing age and career paths of the workforce, and of methodological issues in absence research.     

Statistical inference for the effect of magnetic brain stimulation on a motoneurone

A stochastic model is developed for the possible excitatory and inhibitory effects of a stimulus to the brain on the activity of single human motoneurones. The model consists of a Wiener process for the build-up of underlying potential, a deterministic effect due to the stimulus and a random lag from brain to muscle. Direct likelihood inference for its parameters seems impossible, and we study the use of simulation to estimate the log-likelihood for the parameters of substantive interest. Monte Carlo methods yield point and confidence interval estimates of the membrane excitability underlying excitatory and inhibitory effects. The main qualitative conclusion is that both excitatory and inhibitory effects are unambiguously present. The contribution of statistical analysis to this problem is to provide accurate and apparently reliable inference for the quantities of neurophysiological interest. More generally, our methodology has the potential to make accurate likelihood-based inferences in challenging problems, but the computational burden can be large, particularly if the model is not fully adequate for the data, as in our application.     

Some Effects of the Community on Cultural Integration*

Attitudinal and behavioral data indicative of cultural integration support the hypothesis of decreasing race/ethnic differences. The study data were collected from inmigrant Mexican-Americans and Negroes and a control group of Anglo residents at two points in time separated by eleven years. Tau is used to describe similarity or dissimilarity in race ethnic distributions within each year's data and as a measure of changing race/ethnic response across the eleven-year span.     

Effect of ceasing creatine supplementation while maintaining resistance training in older men

The authors previously found that creatine (Cr) combined with 12 weeks of resistance training enhanced muscle strength and endurance and lean tissue mass (LTM) in older men. Their purpose in this study was to assess these variables with cessation of Cr combined with 12 weeks of reduced training (33% lower volume) in a subgroup of these men (n = 8, 73 years old) compared with 5 men (69 years old) who did not receive Cr. Strength (1-repetition maximum [1-RM]), endurance (maximum number of repetitions over 3 sets at 70-80% 1-RM), and LTM (dual-energy X-ray absorptiometry) were assessed before and after 12 weeks of Cr cessation combined with reduced-volume training. No changes in strength or LTM occurred. Muscle endurance was significantly reduced (7-21%; p <.05), with the rate of change similar between groups. Withdrawal from Cr had no effect on the rate of strength, endurance, and loss of lean tissue mass with 12 weeks of reduced-volume training.