Covariance-regularized regression and classification for high dimensional problems

Summary.  We propose covariance-regularized regression, a family of methods for prediction in high dimensional settings that uses a shrunken estimate of the inverse covariance matrix of the features to achieve superior prediction. An estimate of the inverse covariance matrix is obtained by maximizing the log-likelihood of the data, under a multivariate normal model, subject to a penalty; it is then used to estimate coefficients for the regression of the response onto the features. We show that ridge regression, the lasso and the elastic net are special cases of covariance-regularized regression, and we demonstrate that certain previously unexplored forms of covariance-regularized regression can outperform existing methods in a range of situations. The covariance-regularized regression framework is extended to generalized linear models and linear discriminant analysis, and is used to analyse gene expression data sets with multiple class and survival outcomes.     

Consultants' forum: should post hoc sample size calculations be done?

Pre‐study sample size calculations for clinical trial research protocols are now mandatory. When an investigator is designing a study to compare the outcomes of an intervention, an essential step is the calculation of sample sizes that will allow a reasonable chance (power) of detecting a pre‐determined difference (effect size) in the outcome variable, at a given level of statistical significance. Frequently studies will recruit fewer patients than the initial pre‐study sample size calculation suggested. Investigators are faced with the fact that their study may be inadequately powered to detect the pre‐specified treatment effect and the statistical analysis of the collected outcome data may or may not report a statistically significant result. If the data produces a “non‐statistically significant result” then investigators are frequently tempted to ask the question “Given the actual final study size, what is the power of the study, now, to detect a treatment effect or difference?” The aim of this article is to debate whether or not it is desirable to answer this question and to undertake a power calculation, after the data have been collected and analysed. Copyright © 2008 John Wiley & Sons, Ltd.     

Parental characteristics and the schooling progress of the children of immigrant and U.S.-born blacks

In this study, I examine disparities in schooling progress among children born to immigrant and U.S.-born blacks. I find that in one- and two-parent families, children born to black immigrants are less likely to fall behind in school than those born to U.S.-born blacks. In two-parent immigrant families, children born to two immigrant parents have a significant schooling advantage over children born to one immigrant parent. While children born to two immigrant parents in the wealthiest black immigrant families do better in the second generation than in the first, the reverse is observed among children in less wealthy families. These findings contribute in two ways to our understanding of the assimilation processes of children born to black immigrant parents. First, they show that there is a positive association between the number of immigrant parents in a family and children’s schooling performance. Second, they suggest that disparities in the assimilation patterns of the children of black immigrants are a likely product of the interaction between their parental characteristics and the socioeconomic circumstances of their families.     

Determinants of business success: an examination of Asian-owned businesses in the USA

Using confidential microdata from the US Census Bureau, we investigate the performance of Asian-owned businesses. Using regression estimates and a special non-linear decomposition technique, we explore the role that class resources, such as financial capital and human capital, play in contributing to the relative success of Asian businesses. We find that Asian-owned businesses are more successful than white-owned businesses for two main reasons—Asian owners have high levels of human capital and their businesses have substantial start-up capital. Using detailed information on both the owner and the firm, we estimate the explanatory power of several additional factors.

Racial Disparities in Early Criminal Justice Involvement

Criminologists have long reported the existence of racial disparity in the criminal justice system, but the important question is why. While some argue that observed differences are a consequence of more criminal behavior among minorities, the weight of the evidence indicates that this is but a partial explanation. In this paper, we study data from a sample of juveniles to examine how racial differences in early police contact, and important social environments—family, school, and neighborhoods—affect later contact and arrests, controlling for self-reported delinquency. We find that early (in middle school) contact with police is an important predictor of later (high school) arrests. Also we found that, in addition to being male and living in a low-income family, children who have parents who have a history of arrest, who have experienced school disciplinary actions, who have delinquent peers, and who are in networks with deviant adults are more likely to have problems with law enforcement. These factors help to explain racial differences in police contacts and arrests.     

Public Sociology on Twitter: a Space for Public Pedagogy?

Research has examined various elements of Twitter; however, no scholarship has explored how sociologists currently use the platform. This empirically driven paper explores how individuals that self-identify as sociologists on Twitter use the popular micro-blogging social media site. A total of 152,977 tweets from Twitter profiles of 130 sociologists were collected and examined using qualitative media analysis. The potential use of Twitter allows the sociologist to become both the generator and interlocutor of dialogue with publics. We frame our data analysis and discussion around the core theme of expertise - namely, the role that expertise plays in the use of Twitter by sociologists. Our findings indicate that when sociologists used Twitter as sociologists (i.e., drawing upon their stated research expertise) little direct engagement with publics occurred. Thus, while sociologists appear to be using Twitter as a space for public sociology, the use of this interactive platform is mostly limited to the generation of content, a finding consistent with Burawoy’s traditional form of public sociology. Suggestions for future research are noted.     

Fathers' Investments of Money and Time Across Residential Contexts

Fathers' roles in family life have changed dramatically over the past 50 years. In addition to ongoing breadwinning responsibilities, many fathers are now involved in direct caregiving and engagement with children. Yet there is considerable variation in what fathers do, especially depending on whether they live with or away from their child. In this article, the authors use data from the Fragile Families and Child Wellbeing Study (N = 3,869) to describe how fathers' economic capacities (money) and direct involvement with children (time) are associated over child ages 1 to 9 for resident versus nonresident fathers, net of confounding factors. They found suggestive evidence that money and time investments operate differently across residential contexts: Resident fathers experience a trade‐off between market work and time involved with children. In contrast, nonresident fathers' higher economic capacities are associated with more time involvement, underscoring the greater challenge for such fathers to remain actively involved.     

Is ethnic prejudice declining in Britain? Change in social distance attitudes among ethnic majority and minority Britons

Most literature on racial prejudice deals with the racial attitudes of the ethnic majority and ethnic minorities separately. This paper breaks this tradition. We examine the social distance attitudes of white and non‐white British residents to test if these attitudes follow the same trends over time, whether they are driven by the same social processes and whether they are inter‐related. We have three main findings. Firstly, social distance from other ethnic groups has declined over time for both white and ethnic minority Britons. For the white majority there are both period and cohort elements to this decline. Secondly, we see some evidence that social distance between the majority and minority groups is reciprocal. Specifically, minorities who experience rejection by the white British feel a greater sense of distance from them. Thirdly, we find that all groups share the perception of the same ethnic hierarchy. We see evidence of particularly widespread hostility towards Muslim Britons from all ethnic groups suggesting that Muslims are singled out for negative attention from many British residents of all other backgrounds, including a large number who do not express hostility to other groups.     

Grandparents' Education and Infant Health: Pathways Across Generations

Using data from the Fragile Families and Child Wellbeing survey linked to respondents' medical records (N = 2,870), this study examines the association between grandparents' education and birth outcomes and explores potential pathways underlying this relationship. Results show that having a grandfather with less than a high school education was associated with a 93 gram reduction in birth weight, a 59% increase in the odds of low birth weight, and a 136% increase in the odds of a neonatal health condition when compared with having a grandfather with a high school education or more. These associations were partially accounted for by mother's educational attainment and marital status as well as by prenatal history of depression, hypertension, and prenatal health behaviors, depending on the specific outcome. The findings from this study call for heightened attention to the multigenerational influences of educational attainment for infant health.     

Familial Patterning and Prevalence of Male Androphilia in Samoa

Previous research established that male androphilia (i.e., sexual arousal and attraction to adult males) clusters in families. Some studies find that male androphilia clusters in both the paternal and maternal lines, while others find that it clusters only in the latter. Most of the research investigating the familial nature of male androphilia has taken place in Western cultural contexts that are problematic for such research because they are characterized by low fertility. To address this, our previous work has examined familial patterning of male androphilia in Samoa, a high-fertility population in which androphilic males are readily identified due to their public status as fa’afafine (a third gender category). Building on this work, the present study replicated the familial nature of male androphilia in Samoa using a sample size that was ~122% larger than the one we previously employed (N = 382, M ±SD age: 29.72 years ±10.16). Samoan fa’afafine had significantly more fa’afafine relatives in their maternal and paternal lines compared to Samoan gynephilic males (p < .001). The prevalence of male androphilia was equivalent across both the paternal and maternal lines (all p > .15). The revised prevalence estimate of male androphilia in Samoa falls between 0.61% and 3.51%.