Search Results

This chapter has two main goals: to produce data that provide a complete picture of the preferences of insiders, outsiders, and upscale groups; and to test whether these preferences fit into the partisanship model proposed in this book. The chapter proceeds as follows. The first section provides a brief explanation of the survey used in the analysis and the way insiders, outsiders, and upscale groups have been categorized. The second section offers a detailed explanation of the individual preferences implied in the insider-outsider model and an initial and descriptive assessment of their accuracy. The third section contains a systematic multilevel analysis of the individual preferences of insiders, outsiders, and upscale groups. The fourth section introduces the two macro factors which, in Chapter 2, were hypothesized to affect the differences between insiders and outsiders: job security and corporatism. The final section presents multilevel maximum likelihood models estimating the effects of job security and corporatism. The results corroborate the model's claims: lower levels of employment protection do indeed make insiders more like outsiders (i.e., more supportive of labour market policy). The results support an economic insider-outsider interpretation of the effects of corporatism on insider preferences.

This chapter introduces the multilevel model for change, demonstrating how it allows researchers to address within-person and between-person questions about change simultaneously. Although there are several ways of writing the statistical model, a simple and common approach is adopted that has much substantive appeal. It specifies the multilevel model for change by simultaneously postulating a pair of subsidiary models—a level-1 submodel that describes how each person changes over time, and a level-2 model that describes how these changes differ across people.

This chapter demonstrates how to apply the multilevel model to complex data sets. Section 5.1 begins by illustrating what to do when the number of waves is constant but their spacing is irregular. Section 5.2 illustrates what to do when the number of waves per person differs as well; it also discusses the problem of missing data, the most common source of imbalance in longitudinal work. Section 5.3 demonstrates how to include time-varying predictors in your data analysis. Section 5.4 concludes by discussing why and how to adopt alternative representations for the main effect of TIME.

The previous chapters emphasized the fixed effects in the multilevel model for change. This chapter, in contrast, focuses on the model's random effects as embodied in its error covariance structure. Section 7.1 begins by reviewing the “standard” multilevel model for change, expressed in composite form. Section 7.2 examines this model's random effects, demonstrating that the composite error term is indeed both heteroscedastic and autocorrelated, as preferred for longitudinal data. Section 7.3 compares several alternative error covariance structures and provide strategies for choosing among them.

This chapter reviews the findings of this book and discusses their implications for the study of voter behavior. It argues that the effect of political institutions on voter behavior is typically indirect and contingent and that the most promixate contextual influences are typically different dimensions of the electoral supply in the form of party system polarization and fragmentation.

Intergenerational data necessarily reflect the time and place that the different generations of participants were living in. This chapter aims first to introduce simple concepts to provide an understanding of the founding assumptions and principles, before moving on to more complex analytic methods. As the objectives of analyses may vary substantially across intergenerational studies, there is no easy guideline for analyses, except perhaps that some a priori clarity on the main associations of interest is crucial. Since parents and their offspring are genetically related, intergenerational studies are to some extent genetically informative even if no genotyping was performed. Much of the analyses are concerned with identifying or unravelling the relationship between outcomes and genetic and environmental factors. Ways of handling missing data as well as approaches to deal with non-paternity are also discussed. Illustrative examples are drawn from the two cohort studies.

This chapter introduces a statistical approach for analyzing nested data structures that both accounts for the dependence of observations due to hierarchical arrangements and allows for testing hypotheses at multiple levels. The most common application of multilevel models is for analyses of objects (e.g., people) nested within groups or clusters of some sort. Multilevel models can also be applied to longitudinal data analyses such that the “levels” do not refer to objects nested within groups but instead refer to multiple measurements (e.g., measures made at different occasions/time points) nested within individuals. The chapter illustrates some of the major considerations and basic steps for performing multilevel analyses so that the reader can begin to imagine how to apply this technique to the reader’s own research questions.

The growth of semiparametric regression modeling through generalized estimating questions (GEE) is one of the most influential recent developments in statistical practice. GEE methods are attractive both from a practical and theoretical perspective; they are easy to use, flexible, and make relatively weak assumptions about the distribution of the response of interest. They are closely linked to multilevel models and are commonly regarded as robust relatives of the linear mixed model characterized by Hedeker et al. Because of longstanding tensions existing between two different schools of statistical thought, some who handle longitudinal data may rely either on multilevel models or GEE but not both. The authors see them as complementary instead of referring to the two as rivals.

Linear mixed models, also termed hierarchical linear models (HLM), have been particularly useful for researchers analyzing longitudinal data, but they are not appropriate for all types of longitudinal data. For example, these methods are not able to estimate changes in slope between an outcome variable and potentially time-varying covariates over time. The functional multilevel modeling technique proposed in this chapter addresses this issue by elaborating the linear mixed model to permit coefficients, both random and fixed, to vary nonparametrically over time. Estimation of time-varying coefficients is achieved by adding a local linear regression estimation procedure to the traditional linear mixed model. The main motivation for the current research was methodological challenges faced by drug-use researchers on how to model intensive longitudinal data.

In this chapter, the authors challenge the field to overcome its focus on internal states and behavioral precursors as a substitute for behavior, and offer instead a method for studying behavior in everyday contexts. Using marital conflict as a specific instantiation of an important social behavior, the authors describe the utility of daily diaries as a tool for the assessment of these behaviors. They argue that studying variability in behavior is as essential to understanding behavior as studying mean levels, and offer possibilities for statistical analysis of diary data that focus on such variability. The authors report a reanalysis of couples’ diary data originally published in Bolger et al. (1989) that focuses on questions of variability in marital conflict and reactions to those conflicts. The authors describe how this analysis of variability led to insight about couple conflict that was not apparent from their original analyses.

Repeat measures of the same exposure over time offer a unique opportunity to study the role of developmental trajectories and age-related patterns of function on aspects of health and wellbeing in later life. This chapter illustrates and discusses the interpretation of commonly used methods for relating repeat measures of an exposure to a single later outcome, using an example analysis of life course body mass index and later adult systolic blood pressure. It describes simple multivariable regression models, including the life course plot and conditional change models, relating the outcome to the observed exposures. It then considers growth trajectory approaches where first, a multilevel model or latent class analysis is used to model the repeat exposure data and second, aspects of this model are used as explanatory variables in a regression model with the outcome. These steps can be carried out separately, or as part of one multivariate or structural equation model.

Most studies evaluating relationships between neighborhood characteristics and health neglect to examine and account for the spatial dependency across neighborhoods, that is, how neighboring areas are related to each other, although the possible presence of spatial effects (e.g., spatial dependency, spatial heterogeneity) can potentially influence the results in substantial ways. This chapter first discusses the concept of spatial autocorrelation and then provides an overview of different spatial clustering methods, including Moran’s I and spatial scan statistics as well as different models to map spatial data, for example, spatial Bayesian mapping. Next, this chapter discusses various spatial regression methods used in spatial epidemiology for accounting spatial dependency and/or spatial heterogeneity in modeling the relationships between neighborhood characteristics and health outcomes, including spatial econometric models, Bayesian spatial models, and multilevel spatial models.

An established literature in electoral studies maintains that the electoral system and other contextual factors shape the incentive structure for voters, and thereby influences individual electoral behavior. This chapter describes a theoretical framework for differentiating between various types of contextual effects on individual citizens. It also describes the cross-national variation on key contextual variables for nations in the Cross-national Study of Electoral Systems (CSES). Finally, it introduces the chapters of this volume.

This chapter offers an introduction to coupled differential equation models of self-regulating dynamical systems. It then characterizes a method for approximating the parameters of such models and works through an application of this method to self-disclosure and feelings of intimacy on a sample of married couples. The methods used include Local Linear Approximation (LLA) of derivatives and multilevel modeling, to take into account and estimate the individual differences in parameters of the differential equations. The authors chose to use LLA and multilevel modeling, since this provides a straightforward and simple approach to the approximation of parameters of the models. Innovations in the longitudinal data modeling have led to the development of theories that were modeled and tested using differential equations.

This chapter discusses yet another set of methods aimed at discerning structural dynamics and causal mechanisms of ontogenetic changes: the multivariate analysis of longitudinal changes by means of latent growth models (LGM) and multilevel models (MLM). Similar to other methods, LGM and MLM have specific strengths, limitations, and problems. The chapter highlights central characteristics, using recent examples from research conducted for illustration.

Understanding how and why individuals vary is an important aspect of understanding language function. In assessing literacy in deaf readers, we must supplement normative models of functioning with models that take into account how individual differences enhance or detract from skill attainment. This chapter provides a brief case for and description of multilevel models (sometimes known as hierarchical linear models) as a tool to aid research on individual differences. These kinds of models have been applied successfully to understand variability in both hearing and deaf readers. This chapter explains how multilevel models resemble and differ from other commonly applied data analysis techniques, and why they offer a better alternative than those techniques for many applications within deaf education research.

This chapter discusses the philosophy of multilevel modeling and illustrates this development in the case of the heart. It covers cellular models of the heart; connecting to ion pumps and calcium cycling; connecting to ion pumps and calcium cycling; linking to biochemistry; linking to tissues and organs; coronary circulation.