Search

Categorical data are data for which at least one dependent (outcome) variable is nominal or ordinal. Nominal variables, such as gender, ethnicity, religion, and graduation (or successful completion of any course of study), have a limited number of possible values. Categorical variables are generally divided into two major classes: dichotomous (binary) variables, which have two values (e.g., success vs. failure), and polytomous, which have more than two categories. Polytomous variables are further divided into ordered (ordinal) or unordered. Ordinal variables may have a limited number of categories (such as Likert scales) or a potentially unlimited number of categories, usually a count (e.g., number of times a student interrupts a class). Historical treatment of such data normally involved what was termed nonparametric statistics (e.g., chi-square tests of independence), and involved simple data structures (often just two variables). These methods have evolved to include a wide variety of extensions, including latent variables, graphical causal models, and tree-based methods.

Because of the different philosophy of Bayesian statistics, where parameters are random variables and data are considered fixed, the analysis and presentation of results will differ from that of frequentist statistics. Most importantly, the probabilities that a parameter is in certain regions of the parameter space are crucial quantities in Bayesian statistics that are not calculable (or considered important) in the frequentist approach that is the basis of much of traditional statistics. In this article, I discuss the implications of these differences for presentation of the results of Bayesian analyses. In doing so, I present more detailed guidelines than are usually provided and explain the rationale for my suggestions.

Previous meta-analyses highlight the role of executive functions (EF), encompassing working memory updating, task-switching, and inhibitory control, in reading comprehension, but have not established their role in decoding. Decoding is defined as the use of orthographic patterns to access oral pronunciations. According to the dual route model, decoding involves parallel activation of lexical and phonological routes, which places cognitive demands on EF. We used multivariate meta-analyses to examine associations between decoding, assessed via nonword and word reading tasks, and EF across 65 studies involving children and adolescents (162 effect size estimates, 22,806 observations from 10,173 participants, weighted mean age = 8.8 years). Meta-regression analyses examined a broad set of potential moderators of correlational effect size estimates, including variables related to sample characteristics, task features, and study design. Results indicated significant small-to-moderate correlations between EF constructs and decoding tasks (rz ranged from .28 to .34), with little evidence of moderation. The observed associations between EF and decoding skills in children and adolescents appear to be relatively consistent, even when accounting for moderators related to the sample, task, and study design.

Practice-Based Learning and Improvement, a core competency identified by the Accreditation Council for Graduate Medical Education, carries importance throughout a physician's career. Practice-Based Learning and Improvement is cultivated by a critical review of complications, yet methods to accurately identify complications are inadequate. Machine-learning algorithms show promise in improving identification of complications. We compare a manual-supplemented natural language processing (ms-NLP) methodology against a validated electronic morbidity and mortality (MM) database, the Morbidity and Mortality Adverse Event Reporting System (MARS) to understand the utility of NLP in MM review.The number and severity of complications were compared between MARS and ms-NLP of surgical hospitalization discharge summaries among three academic medical centers. Clavien-Dindo (CD) scores were assigned to cases with identified complications and classified into minor (CD I-II) or major (CD III-IV) harm.Of 7774 admissions, 987 cases were identified to have 1659 complications by MARS and 1296 by ms-NLP. MARS identified 611 (62%) cases, whereas ms-NLP identified 670 (68%) cases. Less than one-third of cases (299, 30.3%) were detected by both methods. MARS identified a greater number of complications with major harm (457, 46.30%) than did ms-NLP (P 0.0001).Both a prospectively maintained MM database and ms-NLP review of discharge summaries fail to identify a significant proportion of postoperative complications and overlap 1/3 of the time. ms-NLP more frequently identifies cases with minor complications, whereas prospective voluntary reporting more frequently identifies major complications. The educational benefit of reporting and analysis of complication data may be supplemented by ms-NLP but not replaced by it at this time.

Introduction This study analyzes the relationship between cognitive bias (CB) and harm severity as measured by Clavien-Dindo Scores (CD). Methods A prospectively collected series of 655 severity matched general surgical cases with complications were analyzed. Cases were evaluated for CB and assigned harm scores as defined by CD grade. Potentially mitigating “debiasing” strategies were identified for each bias attribution. Results Among cases with CB, 24% (55/232) were CD(I-II) and 76% (177/232) were CD(III-V). Odds ratio suggests that serious complications occur nearly 60% more frequently when CB is identified. The CBs identified with severe harm were Overconfidence, Commission, Anchoring, Confirmation, and Diagnosis Momentum. Preliminary data on debiasing strategies suggest diagnosis review, linear reasoning and Type II thinking may be relevant in over 85% of complications. Conclusion The incidence of CB is increased in patients sustaining severe harm. Understanding the specific CBs identified and their mitigating debiasing strategies may improve outcomes.

Background: Randomized experiments yield unbiased estimates of treatment effect, but such experiments are not always feasible. So researchers have searched for conditions under which randomized and nonrandomized experiments can yield the same answer. This search requires well-justified and informative correspondence criteria, that is, criteria by which we can judge if the results from an appropriately adjusted nonrandomized experiment well-approximate results from randomized experiments. Past criteria have relied exclusively on frequentist statistics, using criteria such as whether results agree in sign or statistical significance or whether results differ significantly from each other. Objectives: In this article, we show how Bayesian correspondence criteria offer more varied, nuanced, and informative answers than those from frequentist approaches. Research design: We describe the conceptual bases of Bayesian correspondence criteria and then illustrate many possibilities using an example that compares results from a randomized experiment to results from a parallel nonequivalent comparison group experiment in which participants could choose their condition. Results: Results suggest that, in this case, the quasi-experiment reasonably approximated the randomized experiment. Conclusions: We conclude with a discussion of the advantages (computation of relevant quantities, interpretation, and estimation of quantities of interest for policy), disadvantages, and limitations of Bayesian correspondence criteria. We believe that in most circumstances, the advantages of Bayesian approaches far outweigh the disadvantages.

Bayesian statistics is becoming a popular approach to handling complex statistical modeling. This special issue of Evaluation Review features several Bayesian contributions. In this overview, I present the basics of Bayesian inference. Bayesian statistics is based on the principle that parameters have a distribution of beliefs about them that behave exactly like probability distributions. We can use Bayes’ Theorem to update our beliefs about values of the parameters as new information becomes available. Even better, we can make statements that frequentists do not, such as “the probability that an effect is larger than 0 is .93,” and can interpret 95% (e.g.) intervals as people naturally want, that there is a 95% probability that the parameter is in that interval. I illustrate the basic concepts of Bayesian statistics through a simple example of predicting admissions to a PhD program.

Purpose The purpose of the study was to identify the sex-specific characteristics that predict depression among adult women with diabetes. Methods Data from the 2007-2012 National Health and Nutrition Examination Survey in the United States were used to identify the predictors of depression in a large sample of women ages 20 years and older with diabetes (n = 946). Results When extrapolated to almost 9 million women in the United States ≥ 20 years of age with diabetes, 19.0% had depression. Female-specific significant predictors of depression included younger age (< 65 years old), less than high school graduation, self-rated fair or poor health, inactivity due to poor health, and pain that interferes with usual activities. Marital status and diabetes-related factors (years living with diabetes, use of insulin, parent or sibling with diabetes) were not significant predictors of depression in adult women with diabetes. Conclusion When educating and counseling women with diabetes, diabetes educators should be aware that some of the predictors of depression in women with diabetes differ from those of populations that include both sexes. Depression screening, although important for all women with diabetes, should especially be performed among women with female-specific depression predictors.

Objective Depression is common, frequently resistant to antidepressant treatment, and associated with impairments in cognition and everyday functioning. Computerized cognitive training (CCT) paradigms offer potential to improve cognition, mood and everyday functioning, but their effectiveness is not well established. The goal of this article was to conduct a systematic review and meta-analysis to determine the efficacy of CCT in depressive disorders. Method A search was conducted to identify high quality randomized controlled CCT trials per PRISMA guidelines using PsycINFO and MEDLINE with the keywords “Cognitive training” or “Cognitive remediation” or “Cognitive rehabilitation” and “Depression”. 9 randomized trials for depressed adults met inclusion criteria. Effect sizes (Hedge's g) were calculated for key outcome measures of mood symptom severity, daily functioning, and cognition. A 3-level Bayesian hierarchical linear model was used to estimate effect sizes for each domain and study. Publication bias was assessed using Classic Fail Safe N's and homogeneity was evaluated using Q and I 2 indexes. Results Significant small-moderate effects for Symptom Severity (0.43) and Daily Functioning (0.72), and moderate-large effects for Attention (0.67), Working Memory (0.72), and Global Functioning (1.05) were found. No significant effects were found for Executive Functioning or Verbal Memory. Moderator variable analysis revealed decreased effect of CCT with age. Gender and concurrent medication treatment did not affect the results. Limitations Small sample size, short duration, pseudo-specificity, and high heterogeneity for Verbal Memory measures. Conclusions CCT is associated with improvement in depressive symptoms and everyday functioning, though produces inconsistent effects on cognition.

Several authors have suggested the use of multilevel models for the analysis of data from single case designs. Multilevel models are a logical approach to analyzing such data, and deal well with the possible different time points and treatment phases for different subjects. However, they are limited in several ways that are addressed by Bayesian methods. For small samples Bayesian methods fully take into account uncertainty in random effects when estimating fixed effects; the computational methods now in use can fit complex models that represent accurately the behavior being modeled; groups of parameters can be more accurately estimated with shrinkage methods; prior information can be included; and interpretation is more straightforward. The computer programs for Bayesian analysis allow many (nonstandard) nonlinear models to be fit; an example using floor and ceiling effects is discussed here.

In this article, we respond to Wolery’s critique of the What Works Clearinghouse (WWC) pilot Standards, which were developed by the current authors. We do so to provide additional information and clarify some points previously summarized in this journal. We also respond to several concerns raised by Maggin, Briesch, and Chafouleas after they applied the Standards to a single-case design synthesis published in this journal. The overall purpose of this response is to clarify what the Standards are designed to accomplish and to offer our views about future revisions.

Bayesian statistical methods have great potential advantages for the analysis of data from single case designs. Bayesian methods combine prior information with data from a study to form a posterior distribution of information about their parameters and functions. The interpretation of results from a Bayesian analysis is more natural than those from classical methods, and there are interpretations of useful quantities that are not possible in classical statistics, such as the probability that an effect size is small, or is greater than zero, or is large enough to be considered important. They are not based on asymptotic theory, so small sample size is not a problem for inference. These methods are implemented on free software, and are similar to non-Bayesian software, so analysts familiar with frequentist methods for multilevel data should find the transition relatively painless.

This longitudinal study examined if changes in neuropsychological functioning were associated with the trajectory of symptoms related to attention deficit hyperactivity disorder (ADHD) and impairment between preschool and school age.The sample consisted of 3- and 4-year-old children (N=138) who were identified as being at risk for ADHD based on parent and teacher reports. Neuropsychological functioning was measured annually using the NEPSY at four time points (mean ages, 4.19, 5.36, 6.35, and 7.35 years). ADHD symptoms and impairment were assessed with semiannual parent and teacher reports using the ADHD Rating Scale-IV and the Children's Problems Checklist at 10 time points (mean ages at baseline and final assessment, 4.19 and 8.81 years, respectively). Hierarchical linear modeling was used to assess the trajectories of change in neuropsychological functioning and ADHD severity as well as the association of change in neuropsychological functioning with change in ADHD severity over time.Baseline neuropsychological functioning was not significantly associated with the slope of change in ADHD severity. However, the magnitude of change in neuropsychological functioning was linearly associated with the trajectory of ADHD symptom severity and impairment, such that individuals with greater neuropsychological growth over time had a greater diminution of ADHD severity and impairment. Family socioeconomic status at baseline was significantly associated with initial ADHD severity and impairment, but not with change over time.Interventions that enhance neuropsychological functioning at an early age may be beneficial in attenuating long-term ADHD severity and impairment.

The focus of this paper is to describe Bayesian estimation, including construction of prior distributions, and to compare parameter recovery under the Bayesian framework (using weakly informative priors) and the maximum likelihood (ML) framework in the context of multilevel modeling of single-case experimental data. Bayesian estimation results were found similar to ML estimation results in terms of the treatment effect estimates, regardless of the functional form and degree of information included in the prior specification in the Bayesian framework. In terms of the variance component estimates, both the ML and Bayesian estimation procedures result in biased and less precise variance estimates when the number of participants is small (i.e., 3). By increasing the number of participants to 5 or 7, the relative bias is close to 5% and more precise estimates are obtained for all approaches, except for the inverse-Wishart prior using the identity matrix. When a more informative prior was added, more precise estimates for the fixed effects and random effects were obtained, even when only three participants were included. ispartof: Psychological Methods vol:22 issue:4 pages:760-778 ispartof: location:United States status: published

Researchers in the single-case design tradition have debated the size and importance of the observed autocorrelations in those designs. All of the past estimates of the autocorrelation in that literature have taken the observed autocorrelation estimates as the data to be used in the debate. However, estimates of the autocorrelation are subject to great sampling error when the design has a small number of time points, as is typically the situation in single-case designs. Thus, a given observed autocorrelation may greatly over- or underestimate the corresponding population parameter. This article presents Bayesian estimates of the autocorrelation that greatly reduce the role of sampling error, as compared to past estimators. Simpler empirical Bayes estimates are presented first, in order to illustrate the fundamental notions of autocorrelation sampling error and shrinkage, followed by fully Bayesian estimates, and the difference between the two is explained. Scripts to do the analyses are available as supplemental materials. The analyses are illustrated using two examples from the single-case design literature. Bayesian estimation warrants wider use, not only in debates about the size of autocorrelations, but also in statistical methods that require an independent estimate of the autocorrelation to analyze the data.

In an effort to responsibly incorporate evidence based on single-case designs (SCDs) into the What Works Clearinghouse (WWC) evidence base, the WWC assembled a panel of individuals with expertise in quantitative methods and SCD methodology to draft SCD standards. In this article, the panel provides an overview of the SCD standards recommended by the panel (henceforth referred to as the Standards) and adopted in Version 1.0 of the WWC’s official pilot standards. The Standards are sequentially applied to research studies that incorporate SCDs. The design standards focus on the methodological soundness of SCDs, whereby reviewers assign the categories of Meets Standards, Meets Standards With Reservations, and Does Not Meet Standards to each study. Evidence criteria focus on the credibility of the reported evidence, whereby the outcome measures that meet the design standards (with or without reservations) are examined by reviewers trained in visual analysis and categorized as demonstrating Strong Evidence, Moderate Evidence, or No Evidence. An illustration of an actual research application of the Standards is provided. Issues that the panel did not address are presented as priorities for future consideration. Implications for research and the evidence-based practice movement in psychology and education are discussed. The WWC’s Version 1.0 SCD standards are currently being piloted in systematic reviews conducted by the WWC. This document reflects the initial standards recommended by the authors as well as the underlying rationale for those standards. It should be noted that the WWC may revise the Version 1.0 standards based on the results of the pilot; future versions of the WWC standards can be found at http://www.whatworks.ed.gov .

To increase HCV-related support for patients in substance abuse treatment programs, we implemented an on-site staff training in 16 programs throughout the United States. It aimed to increase participants' self-efficacy in assisting patients with their HCV-related needs. Findings indicate that participants' self-efficacy increased both 1- and 3-months post-training, resulting in providers' perceptions that they were better able to support patients regarding HCV. Implementing an engaging and interactive HCV training for social workers and other substance abuse treatment program staff has the potential to increase their HCV knowledge, self-efficacy, and the HCV-related assistance provided to patients both in the short- and longer-term.

Background Vascular depression has been proposed as a unique diagnostic subtype in late life, yet no study has evaluated whether the specified clinical features associated with the illness are jointly indicative of an underlying diagnostic class. Methods We applied latent class analysis to two independent clinical samples: the prospective, cohort design, Neurocognitive Outcomes of Depression in the Elderly (NCODE) study and the 8-week, multicenter, double blind, placebo-controlled Old-Old study. Results A two-class model consisting of vascular and nonvascular depressed patients provided an excellent fit to the data in both studies, χ 2 (6) = 2.02, p = .90 in the NCODE study and χ 2 (6) = 7.024, p = .32 in the Old-Old study. Although all of the proposed features of vascular depression were useful in identifying the illness, deep white matter lesion burden emerged with perfect sensitivity (1.00) and near-perfect specificity (.95), making it the only indicator necessary to determine class membership. Conclusions These findings, replicated across two independent clinical samples, provide the first support for the internal validity of vascular depression as a subtype of late-life depression.

The articles in the previous, special issue of Evidence-Based Communication Assessment and Intervention provided an excellent review of the meta-analysis of single-case designs. This article weaves commentary about those articles into a larger narrative about two major lines of attack on this problem: the use of parametric approaches like regression and multilevel modeling, and the development of parametric and nonparametric effect-size estimators. On each of these two topics, we describe an agenda of research topics that need to be addressed; and we also introduce a new effect-size estimator that may prove to be comparable to the usual standardized mean difference statistics (d) widely used in between-groups analysis. The article ends with observations about ways in which developments in the meta-analysis of single-case designs may have far wider implications than previously appreciated. Source of funding: U.S. Department of Education, Institute of Education Science, Grant # H324U050001-06

Objective Qualitative reviews of late-life antidepressant clinical trials suggest that antidepressant response rates in comparator trials are higher than antidepressant response rates in placebo-controlled trials. No quantitative review has been conducted to test this hypothesis. Methods A meta-analysis was conducted of all published articles in peer-reviewed journals from 1985 to the present to identify randomized clinical trials contrasting antidepressant pharmacotherapy to placebo or an active comparator in late-life depressed outpatients. Sixteen studies (nine comparator trials and seven placebo-controlled trials) were identified. Antidepressant response rates in both placebo-controlled and comparator trials were extracted and submitted for analysis using multilevel meta-analysis procedures. Results The authors found significant variability in antidepressant response rates beyond chance. This variability decreased by 27% when the authors included study type in the model. As expected, antidepressant response rates in comparator trials were significantly higher (60%) than antidepressant response rates in placebo-controlled trials (46%). Conclusion Antidepressant response rates are higher in comparator trials as compared to placebo-controlled trials. These findings have important implications for combined medication and psychotherapy trials that use placebo-controlled medication conditions because the response rates from these conditions are likely to be lower than those from unblinded conditions.

This paper shows how confirmatory factor analysis can be used to test second- (and higher-) order factor models in the areas of the structure of abilities, allometry, and the separation of specific and error variance estimates. In the latter area, an idea of Joreskog's is extended to include a new conceptualization of the estimation of validity. Second-order models are placed in a hierarchy of factor analysis models, which shows how the fit of various models can be compared. The concept of discriminability is introduced to describe a situation in which two models may both be identified, and yet the goodness-of-fit for both will be the same. This problem can usually be avoided by careful design of a study. Several examples are discussed.

This article examines whether relationships between individual characteristics and HIV status can be identified when self-report data are used as a proxy for HIV serotest results. The analyses use data obtained from HIV serotests and face-to-face interviews with 7,256 out-of-treatment drug users in ten sites from 1992 to 1998. Relationships between 17 individual characteristics and HIV status (measured both by self-report and serotest results) were determined by (a) comparing and evaluating the fit of both standard and "nonstandard" loglinear models, and (b) evaluating the fit of marginal homogeneity models. The loglinear analyses showed that HIV serostatus was related to individual characteristics in 38 percent of the relationships. In most cases, the strength of the relationships between HIV status and individual characteristics did not differ when HIV status was measured with self-report data.

Summary judgment in federal courts has been widely regarded as an initially underused procedural device that was revitalized by the 1986 Supreme Court trilogy of Celotex, Anderson, and Matsushita. Some recent commentators believe summary judgment activity has expanded to the point that it threatens the right to trial. We examined summary judgment practice in six federal district courts during six time periods over 25 years (1975–2000), extracting information on summary judgment practice from 15,000 docket sheets in random samples of terminated cases. We found that when we controlled for changes over time in the types of cases being filed, the likelihood that a case contained one or more motions for summary judgment increased before the Supreme Court trilogy, from approximately 12 percent in 1975 to 17 percent in 1986, and has remained fairly steady at approximately 19 percent since that time. The increase prior to the 1986 trilogy and the modest changes subsequent to the trilogy would be unexpected by many legal commentators. Although summary judgment motions have increased over this 25-year period, this increase reflects, at least in part, increased filings of civil rights cases, which have always experienced a high rate of summary judgment motions. Surprisingly, no statistically significant changes over time were found in the outcome of defendants’ or plaintiffs’ summary judgment motions, again after controlling for differences across courts and types of cases. These findings call into question the interpretation that the trilogy led to expansive increases in summary judgment. Our analysis suggests, instead, that changes in civil rules and federal case-management practices prior to the trilogy may have been more important in bringing about changes in summary judgment practice.

Organizational researchers, including those carrying out occupational stress research, often conduct longitudinal studies. Hierarchical linear modeling (HLM; also known as multilevel modeling and random regression) can efficiently organize analyses of longitudinal data by including within- and between-person levels of analysis. A great deal of longitudinal research has been conducted in the context of growth studies in which change in the dependent variable is examined in relation to the passage of time. HLM can treat longitudinal data, including data outside the context of the growth study, as nested data, reducing the problem of censoring. Within-person equation coefficients can represent the impact of Time t − 1 working conditions on Time t outcomes using all appropriate pairs of data points. Time itself need not be an independent variable of interest.

Good quantitative evidence does not require large, aggregate group designs. The authors describe ground-breaking work in managing the conceptual and practical demands in developing meta-analytic strategies for single subject designs in an effort to add to evidence-based practice.

Objective Depressed older adults with executive dysfunction (ED) may respond poorly to antidepressant treatment. ED is a multifaceted construct and different studies have measured different aspects of ED, making it unclear which aspects predict poor response. Meta-analytic methods were used to determine whether ED predicts poor antidepressant treatment response in late-life depression and to determine which domains of executive functioning are responsible for this relationship. Methods A Medline search was conducted to identify regimented treatment trials contrasting executive functioning between elderly responders and nonresponders; only regimented treatment trials for depressed outpatients aged 50 and older were included. Following the most recent PRISMA guidelines, 25 measures of executive functioning were extracted from eight studies. Six domains were identified: cognitive flexibility, planning and organization, response inhibition, selective attention, verbal fluency, and the Dementia Rating Scale Initiation/Perseveration composite score (DRS I/P). Hedge's g was calculated for each measure of executive functioning. A three-level Bayesian hierarchical linear model (HLM) was used to estimate effect sizes for each domain of executive functioning. Results The effect of planning and organization was significantly different from zero (Bayesian HLM estimate of domain effect size: 0.91; 95% CI: 0.32–1.58), whereas cognitive flexibility, response inhibition, selective attention, verbal fluency, and the DRS I/P composite score were not. Conclusion The domain of planning and organization is meaningfully associated with poor antidepressant treatment response in late-life depression. These findings suggest that therapies that focus on planning and organization may provide effective augmentation strategies for antidepressant nonresponders with late-life depression.

Objectives We reanalyzed data from a previous randomized crossover design that administered high or low doses of intravenous immunoglobulin (IgG) to 12 patients with hypogammaglobulinaemia over 12 time points, with crossover after time 6. The objective was to see if results corresponded when analyzed as a set of single-case experimental designs vs. as a usual randomized controlled trial (RCT). Study Design and Settings Two blinded statisticians independently analyzed results. One analyzed the RCT comparing mean outcomes of group A (high dose IgG) to group B (low dose IgG) at the usual trial end point (time 6 in this case). The other analyzed all 12 time points for the group B patients as six single-case experimental designs analyzed together in a Bayesian nonlinear framework. Results In the randomized trial, group A [ M = 794.93; standard deviation (SD) = 90.48] had significantly higher serum IgG levels at time six than group B ( M = 283.89; SD = 71.10) ( t = 10.88; df = 10; P M D = 511.05 [standard error (SE) = 46.98]. For the single-case experimental designs, the effect from an intrinsically nonlinear regression was also significant and comparable in size with overlapping confidence intervals: M D = 495.00, SE = 54.41, and t = 495.00/54.41 = 9.10. Subsequent exploratory analyses indicated that how trend was modeled made a difference to these conclusions. Conclusions The results of single-case experimental designs accurately approximated results from an RCT, although more work is needed to understand the conditions under which this holds.

Objectives. We examined the potential for glycemic control monitoring and screening for diabetes in a dental setting among adults (n = 408) with or at risk for diabetes. Methods. In 2013 and 2014, we performed hemoglobin A1c (HbA1c) tests on dried blood samples of gingival crevicular blood and compared these with paired “gold-standard” HbA1c tests with dried finger-stick blood samples in New York City dental clinic patients. We examined differences in sociodemographics and diabetes-related risk and health care characteristics for 3 groups of at-risk patients. Results. About half of the study sample had elevated HbA1c values in the combined prediabetes and diabetes ranges, with approximately one fourth of those in the diabetes range. With a correlation of 0.991 between gingival crevicular and finger-stick blood HbA1c, measures of concurrence between the tests were extremely high for both elevated HbA1c and diabetes-range HbA1c levels. Persons already diagnosed with diabetes and undiagnosed persons aged 45 years or older could especially benefit from HbA1c testing at dental visits. Conclusions. Gingival crevicular blood collected at the dental visit can be used to screen for diabetes and monitor glycemic control for many at-risk patients.

Hepatitis C virus (HCV) infection is a global health problem, and in many countries (including the U.S.), illicit drug users constitute the group at greatest risk for contracting and transmitting HCV. Drug treatment programs are therefore unique sites of opportunity for providing medical care and support for many HCV infected individuals. This paper determines subtypes of a large sample of U.S. drug-free treatment programs (N=333) according to services they provide to patients with HCV infection, and examines the organizational and aggregate patient characteristics of programs in these subtypes. A latent class analysis identified four subtypes of HCV service provision: a "Most Comprehensive Services" class (13% of the sample), a "Comprehensive Off-Site Medical Services" class (54%), a "Medical Monitoring Services" class (8%) and a "Minimal Services" class (25%). "Comprehensive" services class programs were less likely to be outpatient and private for profit than those in the other two classes. It is of concern that so many programs belong to the "Minimal Services" class, especially because some of these programs serve many injection drug users. "Minimal Services" class programs in the U.S. need to innovate services so that their HCV infected patients can get the medical and support care they need. Similar analyses in other countries can inform their policy makers about the capacity of their drug treatment programs to provide support to their HCV infected patients.

Illegal drug use remains one of the United States' most serious health problems, and the “War on Drugs” continues without an end in sight. Antidrug programs, which offer the potential to reduce substance abuse problems, are a component of efforts to deal with the problem, but they operate absent adequate scientific analysis. Although policy has shifted from a focus on supply control to one that includes prevention and treatment, research and theory lag behind program implementation. Thus, for example, community-based programs designed to change norms and systems of substance use have been widely promoted despite the lack of data to support their use. The present paper summarizes findings from an evaluation of a large national demonstration program, “Fighting Back.” Results of the evaluation of broad-based community initiatives in a dozen communities show that the programs failed to reduce rates of substance use and associated harms. These findings, along with other evidence, place reliance on community-based programs at odds with public rhetoric. To deal more effectively with substance abuse, there is a need to move from “grading” programs to understanding why and how interventions function.

In this article, multilevel latent class analysis is used to examine the structure of heavy alcohol use. A model with three latent classes (types) of people fits the data well: those who seldom suffer major consequences from heavy drinking, those who typically suffer only a small number of major consequences, and those who suffer serious consequences in a large number of areas of their lives. Individuals in the first class tend to be older, female, with domestic partners, more highly educated, employed, and either Catholic, Muslim, or Baptist.

We demonstrate a model for categorical data that parallels the MIMIC model for continuous data. The model is equivalent to a latent class model with observed covariates; further, it includes simple handling of missing data. The model is used on data from a large-scale study of HIV that had both biological measures of infection and self-report (missing on some cases). The model allows the determination of sensitivity and specificity of each measure, and an assessment of how well true HIV status can be predicted from characteristics of the individuals in the study.

Many HIV positive drug users are unaware that they have the virus, either because they never obtained testing for HIV or because they submitted a biological specimen for testing but never returned to obtain the result of the test. Using data collected from a large multi-site sample of out-of-treatment HIV positive drug users (N=1,544), we identify a variety of socio-demographic characteristics and drug use and sexual risk behaviors that differentiate HIV-positive individuals who had obtained HIV testing in the past and those who had not and, among those who had been tested, what differentiates individuals who had returned to obtain their HIV test results and those who had failed to return. Results of the analyses suggest that there is a need to target different subgroups of high risk drug users for interventions to obtain HIV testing as compared with those that need encouragement to obtain the results of this testing.

Infinite parameter estimates in logistic regression are commonly thought of as a problem. This article shows that in principle an analyst should be happy to have an infinite slope in logistic regression, because it indicates that a predictor is perfect. Using simple approaches, hypothesis tests may be performed and confidence intervals calculated even when a slope is infinite. Some functions of parameters that are infinite are still well defined, and reasonable estimates of these quantities (in particular, LD50) may be obtained even when the maximum likelihood estimates do not, in a strict sense, exist. Surprisingly, these techniques can provide more reasonable and useful results than the most popular alternative method, exact logistic regression.

When analyzing a contingency table, the most frequently used procedure is to test the association between variables using a chi-square test of independence. If the variables are not independent, no further testing is typically done. However, the chi-square can be decomposed, or partitioned, into components to test hypotheses concerning the nature of the differences among categories of the variables. One can determine which categories of one variable are (and are not) the same in their relation to the other variable. Keywords: chi-square; decomposition; contingency table; categorical data; partition; nonstandard model

Analysts evaluating the strengths of relationships between variables in behavioral science research must often contend with the problem of missing data. Analyses are typically performed using data for cases that are either complete in all the variables, or assume that the data are missing at random. Often, these approaches yield biased results. Using empirical data, the current work explores the implications and consequences of using various statistical models to describe the association of two variables, one ordinal and one dichotomous, in which data are incomplete for the dichotomous variable. These models explicitly reflect the missing data mechanism; models that hypothesize nonignorable nonresponse are given particular attention. Both the statistical fit and substantive consequences of these models are examined. This new methodological approach to examining nonignorable nonresponse can be applied to many behavioral science data sets containing an ordinal variable.

Many meta-analysts incorrectly use correlations or standardized mean difference statistics to compute effect sizes on dichotomous data. Odds ratios and their logarithms should almost always be preferred for such data. This article reviews the issues and shows how to use odds ratios in meta-analytic data, both alone and in combination with other effect size estimators. Examples illustrate procedures for estimating the weighted average of such effect sizes and methods for computing variance estimates, confidence intervals, and homogeneity tests. Descriptions of fixedand random-effects models help determine whether effect sizes are functions of study characteristics, and a random-effects regression model, previously unused for odds ratio data, is described. Although all but the latter of these procedures are already widely known in areas such as medicine and epidemiology, the absence of their use in psychology suggests a need for this description.

For Latinas with fasting plasma glucose (FPG) levels in the prediabetes and diabetes ranges, early detection can support steps to optimize their health. Data collected in 2009-2010 indicate that 36.7% of Latinas in the United States had elevated FPG levels. Latinas with elevated FPG who were unaware of their diabetes status were significantly less likely than non-Hispanic White and non-Hispanic Black women to have seen a health care provider in the past year (75.8%, 92.9%, and 90.2%, respectively; p = .018). With almost 1 million Latinas in the United States with elevated FPG unaware of their diabetes risk, and less likely than other at-risk women to see health care providers, there is an urgent need to establish alternate sites of opportunity for their diabetes screening.

Children with attention deficit/hyperactivity disorder (ADHD) have poorer neuropsychological functioning relative to their typically developing peers. However, it is unclear whether early neuropsychological functioning predicts later ADHD severity and/or the latter is longitudinally associated with subsequent neuropsychological functioning; and whether these relations are different in children with and without early symptoms of ADHD. This study aimed to examine the longitudinal associations between ADHD severity and neuropsychological functioning among children at high and low risk of developing ADHD. Hyperactive/Inattentive (H/I; n = 140) and Typically developing (TD; n = 76) preschoolers (age 3-4 years) were recruited (BL) and followed annually for 3 years (F1, F2, and F3). Teachers rated the children's ADHD severity and impairment using the Behavior Assessment System for Children-2 and the Children's Problem Checklist, respectively. Parent reports of children's ADHD severity were obtained using the Kiddie-Schedule for Affective Disorders and Schizophrenia - Present and Lifetime version. Neuropsychological functioning was assessed using the NEPSY. In the full sample, there were bidirectional longitudinal associations between neuropsychological functioning and ADHD severity between F1 and F3. Among H/I children, neuropsychological functioning at F1 and F2 predicted ADHD severity at F2 and F3, respectively. In contrast, among TD children the only significant relationship observed was that elevated ADHD symptoms at F2 were associated with poorer neuropsychological functioning at F3. Improved neuropsychological functioning may attenuate ADHD symptoms and associated impairment among H/I children during the early school years. Interventions designed to improve neuropsychological functioning among young H/I children may be beneficial in reducing their ADHD severity.