Publication Type: PosterReview Method: Peer ReviewedAbstract: For the purposes of crime policy and reporting, various types of violent crime are often aggregated, most prominently in the context of the USDOJ's Uniform Crime Reporting system. The violent crime index takes for granted that there is coherence in a given jurisdiction among the discrete crime types that comprise it -- assault, homicide, rape, and robbery, yet studies of crime forecasting techniques and theories suggest that more precise aggregates might be warranted. This poster will look at the validity of the construct of aggravated violence in a medium-sized Northeast city when using risk terrain modeling. Aggravated violence is defined for these purposes as assault, robbery, and aggravated assault. Risk terrain modeling’s predictive validity of aggravated violence will be compared to the individual crime types’ predictive validity to determine whether an aggregate dependent variable is an appropriate approach when running risk terrain models.
Publication Type: Conference Paper/Unpublished ManuscriptReview Method: Peer ReviewedAbstract: Recent technological advancements make it easier for researchers to test complex statistical models. Moderated mediation is one such model, in which a process (mediation) is contingent on another variable (moderation). This tutorial describes proper parameterization of specific types of moderated mediation models: models in which multicategorical variables are used as either independent variables or moderators. This procedure is especially useful for scholars whose research often necessitates multicategorical designs, such as political communication scholars (e.g., conservative, liberal, or independent respondents) and media violence scholars (e.g., participants exposed to violent media, prosocial media, or neutral media). It allows them to test moderated mediation within such designs without altering the nature of their datasets. In other words, previously popular practices such as collapsing conditions and excluding conditions are no longer necessary with the implementation of this procedure.
Publication Type: AbstractAbstract: Objective: To determine if timing of OSCE performance affects overall scores and identify factors associated with intergroup variability in the OSCE scores.
Methods: A 3-station OSCE was conducted in one day for 136 second-year pharmacy students. Station 1 was patient education on proper insulin injection technique. Station 2 was warfarin dosage adjustment and counseling. Station 3 was recommendation to a physician on renal dosage adjustment. The students were divided into 4 groups. Groups 1 and 2 performed the OSCE in the morning, and groups 3 and 4 performed in the afternoon. To address the concern that the afternoon groups may unfairly benefit from sharing of OSCE content by the morning groups, the scores for the 4 groups were analyzed using the student t-test. The content of each OSCE station was also evaluated qualitatively to identify factors that may promote intergroup variability in performance.
Results: Group 4 performed significantly better in station 3 compared to group 1 (12.61±1.28 vs. 10.67±2.07, p<0.001) but performed significantly worse in station 1 (12.87±1.74 vs. 13.97±1.49, p=0.004). There were no significant intergroup differences for station 2. Station 1 required performance of a skill and station 2 required problem solving, while station 3 required communication of a specific knowledge with a correct answer.
Implications: Sharing of OSCE content seems to occur among students but does not always serve favorably for the latter groups. OSCEs requiring skill performance or problem solving minimize advantage to the latter groups even in the presence of information sharing.
Publication Type: Conference Paper/Unpublished ManuscriptReview Method: Peer ReviewedAbstract: Quasi-instrumental variables are instruments that are not perfectly exogenous (Bartels 1991). In this paper, I examine how different instrumental variable estimators are affected by using quasi-instruments instead of true instruments. Using Monte Carlo methods, I explore the properties of 2SLS, LIML, and Jackknife estimators. I find that all estimators are seriously biased and inconsistent. I then use these methods to estimate the effect of spending on electoral success in U.S. Senate elections, using data from Gerber (1998).
Check author's web site for an updated version of the paper.