The Association of Survey Setting and Mode with Self-Reported Health Risk Behaviors among High School Students

This study examined whether the prevalence of self-reportedhealth risk behaviors among high school students varied by surveysetting (school versus home) and mode of administration (paperand pencil versus computer). Students in grades 9 and 11 wereassigned randomly to one of four conditions—school paper-and-pencilinstrument (PAPI), school computer-assisted self-interview (CASI),home PAPI, and home CASI. During the spring of 2004, 4,506 studentscompleted identically worded questionnaires based on the YouthRisk Behavior Survey questionnaire. Logistic regression analysescontrolling for sex, grade, and race/ethnicity revealed thatsetting was associated significantly with the reporting of 30of the 55 risk behaviors examined, and mode was associated significantlywith the reporting of 7 of the 55 behaviors. For all behaviorswith a significant setting main effect, the odds of reportingthe behavior were greater among students who completed questionnairesat school than among students who completed questionnaires athome. For all behaviors with a significant mode main effect,PAPI mode students had lower odds of reporting the behaviorthan CASI mode students. Because social measurement researchassumes that higher prevalence estimates are more valid thanlower estimates, methodological factors shown to increase estimates,such as setting and mode, should be considered when planningsurveys.     

Regression Estimation Using Multivariate Failure Time Data and a Common Baseline Hazard Function Model

Recent ‘marginal’ methods for the regression analysis of multivariate failure time data have mostly assumed Cox (1972)model hazard functions in which the members of the cluster have distinct baseline hazard functions. In some important applications, including sibling family studies in genetic epidemiology and group randomized intervention trials, a common baseline hazard assumption is more natural. Here we consider a weighted partial likelihood score equation for the estimation of regression parameters under a common baseline hazard model, and provide corresponding asymptotic distribution theory. An extensive series of simulation studies is used to examine the adequacy of the asymptotic distributional approximations, and especially the efficiency gain due to weighting, as a function of strength of dependency within cluster, and cluster size. This revised version was published online in July 2006 with corrections to the Cover Date.     

Explaining the social patterns of depression: control and problem solving--or support and talking?

Research on the social patterns of depression in the community finds consistently that high levels of education and income, being male, and being married are associated with lower levels of depression. We attempt to explain these patterns as the result of two essential social perceptions: the sense of controlling one's own life rather than being at the mercy of powerful others and outside forces, and the sense of having a supportive and understanding person to talk to in times of trouble. In theory, the sense of control reduces depression because it encourages active problem solving, and the sense of support reduces depression because it provides others to talk to. We find evidence for the first proposition: persons who feel in control of their lives are more likely to attempt to solve problems. Perceived control and problem solving decrease depression and largely explain the effects of income and education on depression. We find, however, that support has mixed effects. Support decreases depression, but talking to others when faced with a problem, which increases with the level of support, increases depression. Support explains a small part of the effect of marriage on depression. Control and support have an interactive effect on depression, suggesting that control and support can substitute for one another to decrease depression: a high level of one reduces the need for the other, and a low level of one is remedied by a high level of the other.     

The Statistical Analysis of Preference Data

A new statistic and a new method of analysis are proposed for data where a sample of respondents provides a preference ordering of some treatments. The new preference statistic is compared with the Friedman statistic, particularly for an example where 12 home owners each ranked four grasses. The new analysis provides a more natural and less misleading assessment of where the differences occur than an analysis based on the rank sums of the Friedman statistic. The new analysis is also more robust to deviations from the classical location problem, is not related to election methods known to have undesirable characteristics and adheres to the Condorcet criterion for election methods.     

QUALITY CONTROL WITH MULTIVARIATE DATA

The paper examines statistical process control of bivariate and multivariate data, using in particular the multivariate equivalents of the univariate Shewhart chart, CUSUM chart and the Exponentially Weighted Moving Average chart. This illustrates the usefulness of Principal Component methods in statistical process control with multivariate data.