Modellvergleich und Ergebnisinterpretation in Logit- und Probit-Regressionen

In the social sciences logit and probit models are often used multivariate data analysis procedures for binary dependent variables. Both procedures can be thought of as resting on a linear model for an unobserved variable y* from which a nonlinear model for the probability of y?=?1 is derived. We first show that compared to linear models this nonlinearity leads to problems of interpreting results from such analysis. In particular odds ratios (exponentiated logit coefficients) often used in logistic regression are problematic in this respect. Instead we recommend using graphical procedures and reporting (corrected) average marginal effects (AME). Based on a series of Monte-Carlo simulations we next demonstrate that the regression coefficients from logit and probit models should not be compared between nested models. Because model building in the social sciences often employs a stepwise procedure a method allowing valid comparisons of effect sizes between models would be advantageous. Results from our simulation study show that average marginal effects and regression coefficients corrected by a method proposed by Karlson et al. (Sociological Methodology 42, 2012) lead to satisfactory results in many different scenarios. In contrast, y*-standardized coefficients are of limited utility and coefficients from a linear probability model should only be used with normally distributed variables.     

Vulnerable Emotional Expression In Emotion Focused Couples Therapy: Relating Interactional Processes To Outcome

This study examined whether interactions characterized by high expression of emotional vulnerability in one partner followed by a highly supportive response style by the other partner predicted greater improvement on domains of forgiveness, unfinished business, trust, and relationship satisfaction in a sample of 32 couples presenting for Emotion Focused Couples Therapy with unresolved emotional injuries. For each outcome measure, two separate hierarchical regression models were tested (injured partner vulnerability and offending partner supportiveness; offending partner vulnerability and injured partner supportiveness). Both models significantly predicted improvement on the majority of outcome measures. Practice suggestions for working with emotionally injured couples are provided in light of the findings.     

Spectral goodness of fit for network models

We introduce a new statistic, ‘spectral goodness of fit’ (SGOF) to measure how well a network model explains the structure of the pattern of ties in an observed network. SGOF provides a measure of fit analogous to the standard R² in linear regression. Additionally, as it takes advantage of the properties of the spectrum of the graph Laplacian, it is suitable for comparing network models of diverse functional forms, including both fitted statistical models and algorithmic generative models of networks. After introducing, defining, and providing guidance for interpreting SGOF, we illustrate the properties of the statistic with a number of examples and comparisons to existing techniques. We show that such a spectral approach to assessing model fit fills gaps left by earlier methods and can be widely applied.     

A test of uncertainty,expectations, and error response in two strike models

An information-uncertainty form of the Hicks strike model is used to test earlier work on the role of forecasts and uncertainty in determining strike activity. The expected zero coefficients for economic forecasts appear in preferred equations, but the expected positive coefficient for inflation uncertainty often appears as significantly negative. Alternative formulations and reasoning make the results appear somewhat more plausible. The performance of the Hicks model is contrasted with an updated Ashenfelter and Johnson model, which performs and predicts well without any untidy coefficients. Both models predict better than naive forecasting. Certain data and concept refinements are added to the testing of both analyses to bring them closer to the spirit of their models and to established research in wage determination and macroeconomics.     

Conditional Fatherhood: Identity Theory and Parental Investment Theory as Alternative Sources of Explanation of Fathering

Two alternative theoretical models of parenting, identity theory and parental investment theory, are investigated as sources of explanation of men's fathering attitudes and behaviors. Four dimensions of fathering are explored: responsivity, harshness, behavioral engagement, and affective involvement. Concepts from identity theory operationalized as predictors include father role salience, role satisfaction, and reflected appraisals. From parental investment theory, concepts included investment maximization, contingent commitment, and paternity certitude. Using telephone survey data drawn from a community‐based probability sample of 208 fathers, each of the four individual indicators of fathering and a composite fathering measure were regressed against the theoretical predictors in hierarchical regression analyses. Both theoretical models were significant, with identity theory predictors accounting for a greater proportion of variance than the parental investment theory predictors. This study underlines the importance of social psychological variables to understanding variations in men's commitments to children.     

Addressing self-selection effects in evaluations of mutual help groups and professional mental health services: An introduction to two-stage sample selection models

Because random assignment to conditions is often neither possible nor desirable in longitudinal evaluations of mutual help organizations, the influence of self-selection effects must be assessed in order to accurately interpret outcome data. One approach to adjusting for self-selection effects is to control for covariates that predict outcome using statistical procedures such as analysis of covariance (ANCOVA), partial correlations, and hierarchical regression. This approach has considerable power, but is less useful when an evaluator is interested in directly modeling the process of entry into a program and incorporating information on the factors affecting self-selection into estimation of program effects. Two-stage sample selection models are designed to address such situations. These models rely on regression procedures in which program participation is modeled in an initial equation, which yields a sample selection correction factor. The correction factor is included with participation in a second equation that predicts outcome. This two-stage procedure allows the evaluator to interpret the observed effects of a professional service or a self-help group in the context of the magnitude and direction of selection effects. We compare and contrast the covariate control and sample selection models in a longitudinal study of the effects of participation in Alcoholics Anonymous on drinking behavior.     

Marital status and psychological well-being: An individual-environment congruence perspective

The role of the individual's relationship to his environment has achieved central prominence in social science theory, but has not yet been adequately represented and measured through numerical methods. A variety of multivariate regression techniques have been proposed to incorporate individual and environmental level measures in a single regression equation. This article presents a special case of a contextual analysis model in which two sets of predictor variables are statistically treated to create a third set of predictor variables called congruence measures. Each of the three sets of predictor variables are then run separately in multiple regression equations against a fourth set of individual level outcome variables to determine the relative gain in predictive power of the congruence measures over the individual and environmental level variables when they are run alone. Martital status variables are used to illustrate and examine the model.     

Missing data in networks: exponential random graph (p1) models for networks with non-respondents

Survey studies of complete social networks often involve non-respondents, whereby certain people within the “boundary” of a network do not complete a sociometric questionnaire—either by their own choice or by the design of the study—yet are still nominated by other respondents as network partners. We develop exponential random graph (p¹) models for network data with non-respondents. We model respondents and non-respondents as two different types of nodes, distinguishing ties between respondents from ties that link respondents to non-respondents. Moreover, if we assume that the non-respondents are missing at random, we invoke homogeneity across certain network configurations to infer effects as applicable to the entire set of network actors. Using an example from a well-known network dataset, we show that treating a sizeable proportion of nodes as non-respondents may still result in estimates, and inferences about structural effects, consistent with those for the entire network.If, on the other hand, the principal research focus is on the respondent-only structure, with non-respondents clearly not missing at random, we incorporate the information about ties to non-respondents as exogenous. We illustrate this model with an example of a network within and between organizational departments. Because in this second class of models the number of non-respondents may be large, values of parameter estimates may not be directly comparable to those for models that exclude non-respondents. In the context of discussing recent technical developments in exponential random graph models, we present a heuristic method based on pseudo-likelihood estimation to infer whether certain structural effects may contribute substantially to the predictive capacity of a model, thereby enabling comparisons of important effects between models with differently sized node sets.     

REFERENCE‐DEPENDENT PREFERENCES,LOSS AVERSION,AND LIVE GAME ATTENDANCE

We develop a consumer choice model of live attendance at a sporting event with reference‐dependent preferences. The predictions of the model motivate the “uncertainty of outcome hypothesis” (UOH) as well as fans' desire to see upsets and to simply see the home team win games, depending on the importance of the reference‐dependent preferences and loss aversion. A critical review of previous empirical tests of the UOH reveals significant support for models with reference‐dependent preferences, but less support for the UOH. New empirical evidence from Major League Baseball supports the loss aversion version of the model. (JEL L83, D12)     

Outcome Expectancies and Risk Behaviors in Maltreated Adolescents

This study examined positive and negative outcome expectancies for risk behaviors, and their association with engagement in risk behaviors, in a sample of 149 maltreated adolescents. “Outcome Expectancies” are evaluative social cognitions about what will occur as a consequence of one's actions. Risk behaviors and outcome expectancies for substance use, sexual behavior, and delinquency were assessed. In all regression models, positive expectancies were significantly related to risk behaviors while negative expectancies, with one exception, were not significantly related. In three of four regression models, significant interactions were found between positive and negative expectancies in predicting risk behaviors. The nature of the interaction differed by type of risk behavior. Beyond demonstrating associations between outcome expectancies and risk behaviors in a maltreated sample, this paper contributes to the study of social information processing by demonstrating significant interactions between positive and negative expectancies.     

Parental Divorce: A Protection from Later Delinquency for Maltreated Children

Family structure and maltreatment (abuse and neglect) have been identified as predictors of youth delinquency, although the relationship is not clear. This article furthers this research by studying a sample of maltreated children (n = 250) in one Midwest county, and through a multiple regression analysis of many risk factors, the study identified only one significant delinquency variable that made delinquency less likely—children who experience parental divorce. Some established risk factors were surprisingly found not to be predictive of later delinquency: minority race, one-parent families, youth substance abuse, recurrent maltreatment, and youth behind in academic grade level. Implications for the family studies and juvenile justice fields are set forth.     

Regression Analysis of Multivariate Binary Response Variables Using Rasch-Type Models and Finite-Mixture Methods

A model is considered for the regression analysis of multivariate binary data such as repeated-measures data (for example, panel data) or multiple-indicators with measures of some underlying characteristic such as attitude or ability (for example, surveys or tests). The model is related to the usual Rasch model, the usual latent-class model, and other familiar models such as logistic regression. In addition to a regression specification, the model includes parameters that describe heterogeneity not accounted for by the predictors. In contrast to most other approaches, a nonparametric specification of the latent mixing distribution is used, leading to a formulation based on scaled latent classes. We examine the relationship between this model and several other models, give a tractable formulation of the likelihood function and likelihood equations, present an algorithm for maximum-likelihood estimation, and analyze marginal and conditional latent structures. The approach is illustrated with longitudinal data from the German Socioeconomic Panel.     

Assessing BSW Student Direct Practice Skill Using Standardized Clients and Self-Efficacy Theory

Entering and exiting BSW students were compared on self-efficacy and on direct practice skill performance with a standardized client. Self-efficacy was tested as a predictor and as a mediator of skill performance. Ordinary least squares hierarchical regression found BSW education to be predictive of higher skill and higher self-efficacy. After controlling for BSW education, self-efficacy negatively predicted direct practice skill, and mediation was not supported as hypothesized. This study suggests caution in the use of self-efficacy as an outcome measure, advances the use of standardized clients for assessment, and affirms the need for the development of valid and reliable instruments for assessment of direct practice skill in social work education.     

Alcohol and Substance Use in Adolescence and Young Adulthood: The Role of Siblings

Interpersonal relationships within and outside the family have been a central part of alcohol and substance use research. Many studies have focused on the role of parents and peers; fewer studies have focused on siblings. This article examines siblings’ roles in alcohol, tobacco, and other drug (ATOD) use patterns and trajectories in the context of familial and nonfamilial factors across time. First, intraclass correlations were used to examine the degree to which older siblings’ ATOD use was associated with younger siblings’ ATOD use. Second, hierarchical regression analyses were conducted to examine the degree to which individual, parent, sibling, and peer factors over time were associated with adolescents’ and young adults’ ATOD use. It should be noted that developmentally proximal predictors were utilized in these models and within-family replication was also examined. Results demonstrate strong associations between older and younger siblings’ ATOD use. Moreover, the developmentally proximal sibling variables were predictive of younger sibling ATOD use in the context of other variables across all substances. Study findings are discussed in terms of identifying promising and potentially malleable points of intervention for future investigators.     

Childrearing,Self‐Control,and Crime: Additional Evidence*

Gottfredson and Hirschi's model of caregiving to produce high self‐control is examined using survey data from a city population of adults. The data permit assessment of the predictive ability of all three elements of the regimen, separately and in com‐bination. While a combination of the elements of the specified model and some of the individual components prove to be statistically significant predictors of self‐control, the coefficients are only modest ones. In addition, the sequential model of crime production set forth by the theorists proves to be questionable. These findings add to an emerging body of research suggesting that self‐control may be dependent on things other than the childhood caregiving regimen identified in self‐control theory.     

Predictors of substance use and family therapy outcome among physically and sexually abused runaway adolescents

There is a dearth of research that examines the impact of family systems therapy on problems among sexually and/or physically abused youth. Given this void, differential outcome and predictors of substance use change were evaluated for abused, as compared with nonabused, runaway adolescents who were randomly assigned to family therapy or treatment as usual. Abused adolescents reported lower family cohesion at baseline, although both abused and nonabused adolescents showed similar substance use reductions. Utilizing hierarchical linear modeling, we found that substance use changed with change in cohesion over time. These findings link change in family functioning to change in adolescent substance use, supporting fiamily systems theory. Findings suggest that a potent target of intervention involves focus on increasing positive communication interactions.     

Sexual coercion within mixed-sex couples: the roles of sexual motives, revictimization, and reperpetration

Research suggests that a history of childhood sexual abuse, and previous experiences of sexual coercion, may predict sexual coercion victimization and perpetration. More recently, sexual motivation has been found to correlate with both consensual and non-consensual sexual activity. However, sexual motivation has not been examined in association with previous experiences of abuse and sexual coercion. The aim of this study was to investigate childhood sexual abuse, previous sexual coercion experiences, and sexual motives of both partners as possible risk factors for current sexual coercion victimization and perpetration within a sample of 209 mixed-sex couples. This study examined whether power, stress relief, partner pressure, and imposition motives contributed unique variance to the prediction of sexual coercion beyond that accounted for by past childhood sexual abuse and sexual coercion events. Using hierarchical logistic regressions, four predictive models were examined for both male and female sexual coercion perpetration and victimization. Results show that childhood sexual abuse was only a significant predictor of female sexual coercion perpetration, whereas male sexual coercion victimization and perpetration were predicted by sexual coercion victimization and perpetration in previous relationships. Power motives were also significant predictors of sexual coercion perpetration, and imposition was a significant predictor of sexual coercion victimization for both genders.