Understanding increments in model performance metrics

The area under the receiver operating characteristic curve (AUC) is the most commonly reported measure of discrimination for prediction models with binary outcomes. However, recently it has been criticized for its inability to increase when important risk factors are added to a baseline model with good discrimination. This has led to the claim that the reliance on the AUC as a measure of discrimination may miss important improvements in clinical performance of risk prediction rules derived from a baseline model. In this paper we investigate this claim by relating the AUC to measures of clinical performance based on sensitivity and specificity under the assumption of multivariate normality. The behavior of the AUC is contrasted with that of discrimination slope. We show that unless rules with very good specificity are desired, the change in the AUC does an adequate job as a predictor of the change in measures of clinical performance. However, stronger or more numerous predictors are needed to achieve the same increment in the AUC for baseline models with good versus poor discrimination. When excellent specificity is desired, our results suggest that the discrimination slope might be a better measure of model improvement than AUC. The theoretical results are illustrated using a Framingham Heart Study example of a model for predicting the 10-year incidence of atrial fibrillation.     

Approximate Confidence Limits for the Difference between Parameters of Two Pólya Distributions

The problem of calculating approximate confidence limits for the difference between success probability parameters of two Pólya distributions is solved for the first time. We suggest some new methods for determining these approximate confidence limits and consider their application to special cases: namely for the binomial and hypergeometric distributions. The various approximate confidence limits are evaluated and compared.     

Peer modeling and college men's sexually impositional behavior in the laboratory

The purpose of this study was to examine the influence of peer modeling on sexually impositional behavior in the laboratory. Male participants with and without a self‐reported history of sexually aggressive behavior viewed video clips depicting nonaggressive and sexually aggressive behavior and then chose one of the clips to show to a female confederate. Half of the participants were first exposed to a male confederate who showed the sexually aggressive video clip to a female confederate. The other half of the participants were exposed to a male confederate who showed a nonaggressive video clip to a female confederate. Exposure to a male confederate who showed a sexually aggressive video clip to a female was associated with participants' choosing to engage in this same behavior. A self‐reported history of sexually aggressive behavior was also associated with participants' showing the sexually aggressive video clip in spite of believing the effect on the female viewer would be negative.     

Homosexual role separation: Implications for analyzing and modeling the spread of HIV

The analysis of the spread of HIV in homosexual populations must be based on knowledge of all factors which promote or inhibit transmission. Recently, data were reported from the Amsterdam cohort study that imply that sexual role separation and the changes therein should be taken into account in analyzing and modeling the spread of HIV. In this paper, a role separation mixing model is described and linked to data from the Amsterdam cohort. The model allows one to study the impact of homosexual role separation on the spread of HIV. The homosexual men are classified into four sexual subgroups based on the preference for: 1) no anal intercourse; 2) anal insertive only; 3) anal receptive only; 4) both insertive and receptive. The analysis accounts for variability and change in homosexual role behavior and loss to follow up. The results support the conjecture that there are large differences between the risks of receptive and insertive anal intercourse, with the latter carrying only minimal risk. The analysis indicates that the spread of HIV is mostly restricted to subgroup 4, the largest and sexually most active subgroup. In this subgroup, the estimated probability of HIV transmission from an infected insertive partner (in subgroup 2 or 4) to a receptive susceptible (in subgroup 4) is between 1 and 5 percent (assuming a closed cohort).     

The Interpretation of Line of Regard by Prelingually Deaf Children

The ability of 5- to 12-year-old deaf children to utilise and interpret another person's visual line of regard (where the eyes are looking) was studied in four experiments using cartoon faces. The children had little difficulty in determining whether or not a face was looking directly at them. They had more difficulty, however, with more complex tasks requiring them to infer mental states of desire and intention from line of regard and to ignore line of regard when it was inappropriate to attend to this cue. Deaf children raised in a hearing environment appear to have more difficulty with these more complex tasks than hearing children. The results are discussed in terms of the special difficulties facing some deaf children in the development of skills involving utilisation of line of regard and the implications for the development of joint attentional behaviour, theory of mind and dyadic social interaction in deaf and hearing children.     

Predicting Teen Motherhood and Teen Fatherhood: Individual Characteristics and Peer Affiliations

In a longitudinal investigation of the antecedents and pathways of teen parenthood, 475 participants (248 females and 227 males) were followed from Grade 7 through early adulthood. In the first year of the investigation, homogeneous subgroups were identified by clustering participants on five characteristics (i.e., aggression, popularity, academic competence, age, and socioeconomic status). At the same time, peer social networks were identified by the Social Cognitive Map procedure. For each participant, the configuration of peer characteristics was also identified. A high correspondence was found between the individual characteristics and the characteristics of closely affiliated peers. Both teen fatherhood and teen motherhood were predicted by individual configurations and peer configurations (e.g., a combination of high aggression, low academic competence, low popularity, and low family SES). Peer characteristics, race, and family socioeconomic status assumed unique roles in predicting teen motherhood. Growth curve analyses showed that teen-mothers differed from non-mothers in their starting points and developmental trajectories. This study demonstrated that a simultaneous examination of both the individual and his/her peer context yielded important information on teen parenthood.     

Romantic Partners' Contribution to the Continuity of Male and Female Delinquent and Violent Behavior

Using data drawn from the National Longitudinal Study of Adolescent Health (Add Health), this study examined males' and females' criminality in young adulthood with models that considered the associations of both their own past delinquency and their current partners' criminality. Specific models considered the main effects of both previous adolescent delinquency and violence and current opposite sex romantic partners' criminality and violence, as well as the interaction of these with each other and by sex of primary participant. Main effects for adolescent delinquency and partners' criminality were significant; however, the results of the interactions by sex suggest different patterns of results for males and females. Models run separately by sex showed that having a prior delinquent history predicted adult criminality for males but not for females, while being in a relationship with a criminal partner increased the odds of criminal behavior in young adulthood for both males and females. A different pattern of results was found for violence in young adulthood. Having a prior violent history predicted adult violence for males and females. However, being in a relationship with a violent partner increased the odds of violent behavior for females only.     

The influence of riparian corridor width on ant and plant assemblages in northern Sydney,Australia

Protecting riparian corridors is a commonly applied environmental policy in urban landscapes. However, empirical data demonstrating their efficacy for biodiversity conservation outcomes is scarce. In this study we investigated whether riparian corridor width influences the diversity and community structure of ant (Hymenoptera: Formicidae) and vascular plant assemblages therein. Eighteen corridors of differing widths were selected from within the Ku-ring-gai Local Government Area, Sydney Australia. Ants were sampled using pitfall traps positioned within rectangular vegetation transects (30 × 10 m). Both ant and plant species richness, when standardised for sampling effort, were unrelated to riparian corridor width. However, significant compositional differences between sites were evident with increased width up to ~50 m. Narrow corridors contained greater abundances of opportunistic ant species and higher proportions of exotic plants. We interpret this to be the result of the greater disturbance/edge influence derived from adjacent upland urban environments. Site beta diversity did not increase with corridor width, suggesting that the exclusion of novel upslope habitats in narrow riparian corridors is not a significant cause of community assemblage modification at these sites. To minimise the impact of deleterious edge effects on ant fauna, environmental managers should seek to retain riparian corridors wider than ~50 m. However, we observed substantial between-site variability of biotic assemblages, irrespective of corridor width. Therefore we recommend that environmental management practice needs to take a catchment-wide approach and consider other parameters that may contribute to riparian health so as to optimise the protection of riparian biodiversity.