George Leckie

Multilevel Modeling of Social Segregation

The traditional approach to measuring segregation is based upon descriptive, non-model-based indices. A recently proposed alternative is multilevel modeling. The authors further develop the argument for a multilevel modeling approach by first describing and expanding upon its notable advantages, which include an ability to model segregation at a number of scales simultaneously. The authors then propose a major extension to this approach by introducing a simple simulation method that allows traditional descriptive indices to be reformulated within a modeling framework. The multilevel approach and the simulation method are illustrated with an application that models recent social segregation among schools in London, UK.

An Original Stepwise Multilevel Logistic Regression Analysis of Discriminatory Accuracy : The Case of Neighbourhoods and Health

Background and Aim Many multilevel logistic regression analyses of “neighbourhood and health” focus on interpreting measures of associations (e.g., odds ratio, OR). In contrast, multilevel analysis of variance is rarely considered. We propose an original stepwise analytical approach that distinguishes between “specific” (measures of association) and “general” (measures of variance) contextual effects. Performing two empirical examples we illustrate the methodology, interpret the results and discuss the implications of this kind of analysis in public health. Methods We analyse 43,291 individuals residing in 218 neighbourhoods in the city of Malmö, Sweden in 2006. We study two individual outcomes (psychotropic drug use and choice of private vs. public general practitioner, GP) for which the relative importance of neighbourhood as a source of individual variation differs substantially. In Step 1 of the analysis, we evaluate the OR and the area under the receiver operating characteristic (AUC) curve for individual-level covariates (i.e., age, sex and individual low income). In Step 2, we assess general contextual effects using the AUC. Finally, in Step 3 the OR for a specific neighbourhood characteristic (i.e., neighbourhood income) is interpreted jointly with the proportional change in variance (i.e., PCV) and the proportion of ORs in the opposite direction (POOR) statistics. Results For both outcomes, information on individual characteristics (Step 1) provide a low discriminatory accuracy (AUC = 0.616 for psychotropic drugs; = 0.600 for choosing a private GP). Accounting for neighbourhood of residence (Step 2) only improved the AUC for choosing a private GP (+0.295 units). High neighbourhood income (Step 3) was strongly associated to choosing a private GP (OR = 3.50) but the PCV was only 11% and the POOR 33%. Conclusion Applying an innovative stepwise multilevel analysis, we observed that, in Malmö, the neighbourhood context per se had a negligible influence on individual use of psychotropic drugs, but appears to strongly condition individual choice of a private GP. However, the latter was only modestly explained by the socioeconomic circumstances of the neighbourhoods. Our analyses are based on real data and provide useful information for understanding neighbourhood level influences in general and on individual use of psychotropic drugs and choice of GP in particular. However, our primary aim is to illustrate how to perform and interpret a multilevel analysis of individual heterogeneity in social epidemiology and public health. Our study shows that neighbourhood “effects” are not properly quantified by reporting differences between neighbourhood averages but rather by measuring the share of the individual heterogeneity that exists at the neighbourhood level.

The role of maternal factors in sibling relationship quality: a multilevel study of multiple dyads per family

BACKGROUND   Although many children grow up with more than one sibling, we do not yet know if sibling dyads within families show similarities to one another on sibling affection and hostility. In the present study the hypotheses were tested that (a) there will be significant between family variation in change in sibling affection and hostility and (b) this between family variation will be explained by maternal affective climate, operationalized as positive and negative ambient parenting, differential parenting and maternal malaise. METHODS   A general population sample of families with single and multiple sibling dyads were visited twice, 2 years apart. Up to 2 children in a family acted as informants; 253 relationships were rated in 118 families. A cross-classified, multilevel model was fit to separate between-family and within-family variance in sibling relationships while simultaneously controlling for informant and partner influences. RESULTS   Thirty-seven percent of the variance in change in sibling affection and 32% of the variance in change in sibling hostility was between family variance. The measured maternal affective climate including, maternal malaise and maternal ambient and differential hostility and affection explained between family differences. CONCLUSIONS   Sibling relationship quality clusters in families and is partly explained by maternal affective climate.

Modeling heterogeneous variance–Covariance components in two-level models

Applications of multilevel models to continuous outcomes nearly always assume constant residual variance and constant random effects variances and covariances. However, modeling heterogeneity of variance can prove a useful indicator of model misspecification, and in some educational and behavioral studies, it may even be of direct substantive interest. The purpose of this article is to review, describe, and illustrate a set of recent extensions to two-level models that allow the residual and random effects variance–covariance components to be specified as functions of predictors. These predictors can then be entered with random coefficients to allow the Level-1 heteroscedastic relationships to vary across Level-2 units. We demonstrate by simulation that ignoring Level-2 variability in residual variances leads the Level-1 variance function regression coefficients to be estimated with spurious precision. We discuss software options for fitting these extensions, and we illustrate them by reanalyzing the classic High School and Beyond data and two-level school effects models presented by Raudenbush and Bryk.

The Impact of Neighbourhood on the Income and Mental Health of British Social Renters

This paper examines the impact of neighbourhood on the income and mental health of individuals living in social housing in the UK. It exploits a dataset that is representative and longitudinal to match people to their very local neighbourhoods. Using this, the paper examines the effect of living in a neighbourhood in which the population is more disadvantaged on the levels and change, over a 10-year window, of income and mental health. It is found that social renters who live with the most disadvantaged individuals as neighbours have lower levels of household income and poorer mental health. However, neighbourhood appears to have no impact on changes in either household income or individual mental health.

Food sharing networks in lowland Nicaragua: An application of the social relations model to count data

Abstract Previous research on food sharing in small-scale societies provides support for multiple evolutionary hypotheses, but evolutionary anthropologists have devoted relatively little attention to the broader relational context of inter-household transfers of food. The present research observes transfers of meat over a yearlong period among 25 households of indigenous Mayangna and Miskito horticulturalists in Nicaragua. To analyze these data, we extend the multilevel formulation of the social relations model to count data, namely the number of portions of meat exchanged between households. Along with other covariates, we examine the effect of an “association index,” which reflects the amount of time that households interact with one another. The association index exhibits a positive effect on sharing, and our overall results indicate that food sharing networks largely correspond to kin-based networks of social interaction, suggesting that food sharing is embedded in broader social relationships between households. We discuss possible extensions of our methodological approach, as appropriate for research on food sharing and social network analysis more broadly.

Individual- and area-level influence on suicide risk: a multilevel longitudinal study of Swedish schoolchildren

BACKGROUND Characteristics related to the areas where people live have been associated with suicide risk, although these might reflect aggregation into these communities of individuals with mental health or social problems. No studies have examined whether area characteristics during childhood are associated with subsequent suicide, or whether risk associated with individual characteristics varies according to childhood neighbourhood context. METHOD We conducted a longitudinal study of 204,323 individuals born in Sweden in 1972 and 1977 with childhood data linked to suicide (n = 314; 0.15%) up to age 26-31 years. Multilevel modelling was used to examine: (i) whether school-, municipality- or county-level characteristics during childhood are associated with later suicide, independently of individual effects, and (ii) whether associations between individual characteristics and suicide vary according to school context (reflecting both peer group and neighbourhood effects). RESULTS Associations between suicide and most contextual measures, except for school-level gender composition, were explained by individual characteristics. There was some evidence of cross-level effects of individual- and school-level markers of ethnicity and deprivation on suicide risk, with qualitative interaction patterns. For example, having foreign-born parents increased the risk for individuals raised in areas where they were in a relative minority, but protected against suicide in areas where larger proportions of the population had foreign-born parents. CONCLUSIONS Characteristics that define individuals as being different from most people in their local environment as they grow up may increase suicide risk. If robustly replicated, these findings have potentially important implications for understanding the aetiology of suicide and informing social policy.

Measures of clustering and heterogeneity in multilevel Poisson regression analyses of rates/count data

Multilevel data occur frequently in many research areas like health services research and epidemiology. A suitable way to analyze such data is through the use of multilevel regression models. These models incorporate cluster‐specific random effects that allow one to partition the total variation in the outcome into between‐cluster variation and between‐individual variation. The magnitude of the effect of clustering provides a measure of the general contextual effect. When outcomes are binary or time‐to‐event in nature, the general contextual effect can be quantified by measures of heterogeneity like the median odds ratio or the median hazard ratio, respectively, which can be calculated from a multilevel regression model. Outcomes that are integer counts denoting the number of times that an event occurred are common in epidemiological and medical research. The median (incidence) rate ratio in multilevel Poisson regression for counts that corresponds to the median odds ratio or median hazard ratio for binary or time‐to‐event outcomes respectively is relatively unknown and is rarely used. The median rate ratio is the median relative change in the rate of the occurrence of the event when comparing identical subjects from 2 randomly selected different clusters that are ordered by rate. We also describe how the variance partition coefficient, which denotes the proportion of the variation in the outcome that is attributable to between‐cluster differences, can be computed with count outcomes. We illustrate the application and interpretation of these measures in a case study analyzing the rate of hospital readmission in patients discharged from hospital with a diagnosis of heart failure.

Orthodontic pain trajectories in adolescents: Between-subject and within-subject variability in pain perception.

INTRODUCTION The objective of this study was to assess the effects of age, sex, and the age-sex interaction on mean pain trajectories and individual variations in the pain experienced by adolescents after orthodontic separator placement. METHODS We included 115 subjects (mean age, 14.99 years; SD, ±1.90 years; 56 boys, 48.7%; 59 girls, 51.3%) in this study. Orthodontic separators were placed in the mesial and distal contact points of the maxillary and mandibular first molars. A 100-mm visual analog scale was used for pain assessment at 11 prespecified times: 1 hour and 2, 4, 12, 24, 36, 48, 72, 96, 120, and 144 hours. A mixed-effects location scale model was used for the data analysis to directly model between-subject and within-subject variabilities in pain in addition to the usual modeling of mean pain as a function of age, sex, and time. RESULTS Mean initial pain after 1 hour of separator placement for the 12- to 15-year-old male group was 13.52 mm on the visual analog scale, which initially increased rapidly (linear estimate, 9.16; P = 0.000; 95% confidence interval [CI], -8.65 to 9.67) but decelerated with time (quadratic estimate, -0.95; P = 0.000; 95% CI, -1.0 to -0.90), suggesting an inverted U-shaped mean pain trajectory. Age, sex, and age-sex interaction effects did not significantly influence initial pain. Compared with the 12- to 15-year-old male group, the 15- to 18-year-old female group reported the steepest rise in pain (estimate, 8.55; P = 0.00; 95% CI, 7.40 to 9.70) and, as a result, experienced the most overall pain. The 12- to 15-year-old male group reported minimum between-subjects variations (SD, ±4.6 mm) as well as within-subjects variations (SD, ±5.5 mm). The between-subjects variations were highest for the 12- to 15-year-old female group (SD, ±9.8 mm), whereas the within-subjects variations were highest for the 15- to 18-year-old female group (SD, ±10.1 mm). CONCLUSIONS The 12- to 15-year-old boys reported the lowest mean average pain intensity and a minimum subjective variation in between-subject and within-subject variances. The 15- to 18-year-old girls experienced maximum mean pain intensity and the highest daily fluctuations in pain intensity. The 12- to 15-year-old girls were the most different from one another in their overall pain experience.

Observed sensitivity during family interactions and cumulative risk: a study of multiple dyads per family

The present study sought to investigate the family, individual, and dyad-specific contributions to observed cognitive sensitivity during family interactions. Moreover, the influence of cumulative risk on sensitivity at the aforementioned levels of the family was examined. Mothers and 2 children per family were observed interacting in a round robin design (i.e., mother-older sibling, mother younger-sibling and sibling-dyad, N = 385 families). Data were dyadic, in that there were 2 directional scores per interaction, and were analyzed using a multilevel formulation of the Social Relations Model. Variance partitioning revealed that cognitive sensitivity is simultaneously a function of families, individuals and dyads, though the importance of these components varies across family roles. Cognitive sensitivity for mothers was primarily attributable to individual differences, whereas cognitive sensitivity for children was predominantly attributable to family and dyadic differences, especially for youngest children. Cumulative risk explained family and individual variance in cognitive sensitivity, particularly when actors were older or in a position of relative competence or authority (i.e., mother to children, older to younger siblings). Overall, this study demonstrates that cognitive sensitivity operates across levels of family organization, and is negatively impacted by psychosocial risk. (PsycINFO Database Record