Statistical Inference After Model Selection

Conventional statistical inference requires that a model of how the data were generated be known before the data are analyzed. Yet in criminology, and in the social sciences more broadly, a variety of model selection procedures are routinely undertaken followed by statistical tests and confidence intervals computed for a “final” model. In this paper, we examine such practices and show how they are typically misguided. The parameters being estimated are no longer well defined, and post-model-selection sampling distributions are mixtures with properties that are very different from what is conventionally assumed. Confidence intervals and statistical tests do not perform as they should. We examine in some detail the specific mechanisms responsible. We also offer some suggestions for better practice and show though a criminal justice example using real data how proper statistical inference in principle may be obtained.     

Examination of an interventionist-led HIV intervention among criminal justice-involved female prisoners

The purpose of this study was to examine the implementation, adherence and protocol fidelity for the Reducing Risky Relationships for HIV (RRR-HIV) study. The RRR-HIV study is a phase III trial of a randomized intervention to reduce human immunodeficiency virus (HIV) risk behaviors among incarcerated women in four US states: Connecticut, Delaware, Kentucky and Rhode Island. The intervention consists of five interventionist-led prison-based group sessions and a sixth individual community-based session. Data on adherence, implementation, acceptability and fidelity of the intervention were obtained from forms completed after the five prison-based sessions by both the interventionist and participant. Data from the sixth session were collected by the interventionist. Of the 363 women recruited to date, 173 (47.6%) have been randomly allocated to the experimental RRR intervention, of which implementation measures were available for 162 (93.6%). Almost three-quarters of women attended all five sessions, each of which lasted a median of 90 minutes, indicating successful implementation of the protocol across multiple study sites. Interventionists and participants alike reported that all of the topics for each session were discussed, suggesting adherence to the protocol. In addition, protocol interventionists indicated that more than 95% of the women were engaged/involved, interested, and understood the materials presented, indicating high levels of acceptability among the participants and fidelity to the intervention protocols. The majority of participants also answered all of the post-test questions correctly, which is another strong indicator of the fidelity to the intervention. Results suggest that the RRR-HIV study has been successfully implemented across multiple study sites. Adherence to the protocol, as well as protocol fidelity and acceptability, were also strong, which is essential to establish prior to examining outcome data.     

Exploring the “Lost and Found” of Justice Theory and Research

Scientific interest in the nature of how people think about justice and fairness began approximately 70 years ago with Stouffer’s classic study on the American soldier. Since then there have been numerous theoretical frameworks and thousands of research studies conducted on what people perceive as fair and the consequences of making a fairness judgment. The goal of this article is to dig through the “lost and found” box of justice research in an attempt to re-examine where we have been, issues and ideas we may have forgotten, and to gain insight on directions we may want to go in the future. The key rediscovery of this review is that perspective matters. Specifically, how people interpret fairness depends critically on whether they are viewing a situation in terms of their material, social, or moral needs and goals. The implications of adopting a contingent theory of how people reason about fairness are discussed.

Risk Factor Models for Adolescent Verbal and Physical Aggression Toward Fathers

Using a French-Canadian population-based longitudinal data set, we examine the impact of socioeconomic factors (paternal education and family structure); inherent individual factors (child gender and developmental trajectories of physical aggression from early to later childhood, problematic substance use), family environment (concurrent parent-child involvement, parental problematic substance use), and prospective and concurrent parenting process variables (mean parental supervision at puberty, concurrent punishment practices) as predictors of adolescent-directed aggression against fathers (in the last 6 months). A childhood behavioral pattern characterized by physical aggression showed the highest risk of adolescent-directed verbal and physical aggression toward fathers, regardless of sex. In terms of parental practices, verbal (and not corporal) punishment in the last 6 months significantly predicted aggression toward fathers. A childhood life-course of violence is likely to culminate in aggression toward fathers during adolescence. Beyond this risk, it seems that harsh verbal punishment by parents builds up the odds of child-directed aggression against fathers.     

Physical activity and adolescent female psychological development

This study explored the relationship between self-reported past and present physical activity levels and high school adolescent women's self-image, sense of mastery, gender role identity, self-perceived physical ability, and self-perceived attractiveness. The Offer Self-image Questionnaire and the Bem Sex Role Inventory were administered to 149 high school adolescent women who were categorized as either physically active, physically inactive, decreasing in physical activity, or increasing physical activity. Results indicated the physically active group members reported self-image and coping characteristics that were significantly more positive than those reported by the physically inactive group members. They also reported significantly more masculine and feminine characteristics and higher physical ability ratings than inactive group members. The decreased activity group did not differ from any of the other groups. These results suggest that physical activity in adolescence, in comparison with physical inactivity, is associated with healthier emotional expression and emotional control, and higher levels of perceived mastery and coping skills for adolescent women. A decrease in physical activity over the course of adolescence does not necessarily result in psychological distress for adolescent women.Received Ph.D. in counseling psychology from Michigan State University. Major interests are concepts of the self, psychological health, and developmental psychology.Received Ph.D. in sport psychology from The Pennsylvania State University. Current research interests include social cognitions in sport and physical activity and the role of sport and physical activity in adolescent development.     

Evolution of the psycho-legal studies program

This article describes Saleem Shah's role as a mentor and consultant to the author. The author focuses on his influence in the development of the author's research program, examining issues in the police arrest of mentally ill offenders and the need for mental health services for jail detainees.     

How to vote,whether to vote: Strategies for voting and abstaining on congressional roll calls

This paper develops and tests a theory of voting and abstaining on Congressional roll calls. The theoretical model assumes that the voting behavior of legislators is oriented toward reelection, and that constituents vote retrospectively. Among the predictions of the theory are that supporters of a program are more likely to abstain than opponents, that conflicted legislators are more likely to vote on the losing side (but will abstain when the vote is very close), and that indifferent legislators will abstain when votes are not close but trade their votes when the outcome is uncertain. The empirical test is based on a series of votes on appropriations for the Clinch River Breeder Reactor from 1975 to 1982. We estimate a nested logit model of, first, the probability of voting for Clinch River, and second, the probability of abstaining from the vote, conditional on preferences regarding the program. All of the empirical results are consistent with the theoretical predictions, and most are statistically significant by conventional standards. The implication is that the abstention decision, as well as yes or no votes, can be purposive, and that the pattern of abstentions is not random among supporters and opponents.The authors gratefully acknowledge research support from the Brookings Institution and the University of Michigan School of Law, and useful comments on an earlier draft by Randall Calvert, Morris Fiorina, Rodney Fort, Amihai Glazer, Keith Krehbiel, Thomas Romer, Kenneth Shepsle, Rodney Smith, Barry Weingast, the UCI Public Choice Study Group, and the Hoover Workshop on Collective Choice.     

Anatomically detailed dolls and the evaluation of child sexual abuse allegations

Although anatomically detailed dolls are becoming the most commonly used tool in the validation of sexual abuse allegations, their use is not without significant problems. This evaluation of the psychometric properties of the dolls using the existing empirical literature reveals that there is considerable difficulty in the standardization of the dolls, administration and scoring procedures, and training in the doll's use. Furthermore, appropriate norms have not been developed. Although good interobserver reliability has been reported, the construct and criterion-related validity has not been demonstrated consistently. The current lack of an empirically based psychometric foundation does not support the use of the dolls in validation interviews, nor the admissibility of doll-based evidence in legal proceedings.