Simulation-Based Finite and Large Sample Tests in Multivariate Regressions.

In the context of multivariate linear regression (MLR) models, it is well known that commonly employed asymptotic test criteria are seriously biased towards overrejection. In this paper, we propose a general method for constructing exact tests of possibly nonlinear hypotheses on the coefficients of MLR systems. For the case of uniform linear hypotheses, we present exact distributional invariance results concerning several standard test criteria. These include Wilks' likelihood ratio (LR) criterion as well as trace and maximum root criteria. The normality assumption is not necessary for most of the results to hold. Implications for inference are two-fold. First, invariance to nuisance parameters entails that the technique of Monte Carlo tests can be applied on all these statistics to obtain exact tests of uniform linear hypotheses. Second, the invariance property of the latter statistic is exploited to derive general nuisance-parameter-free bounds on the distribution of the LR statistic for arbitrary hypotheses. Even though it may be difficult to compute these bounds analytically, they can easily be simulated, hence yielding exact bounds Monte Carlo tests. Illustrative simulation experiments show that the bounds are sufficiently tight to provide conclusive results with a high probability. Our findings illustrate the value of the bounds as a tool to be used in conjunction with more traditional simulation-based test methods (e.g., the parametric bootstrap) which may be applied when the bounds are not conclusive.

Simulation-Based Finite-and Large-sample Inference Methods in Multivariate Regressions and Seemingly Unrelated Regressions

In the context of multivariate regression (MLR) and seemingly unrelated regressions (SURE) models, it is well known that commonly employed asymptotic test criteria are seriously biased towards overrejection. in this paper, we propose finite-and large-sample likelihood-based test procedures for possibly non-linear hypotheses on the coefficients of MLR and SURE systems.

German Validation of the Conners Adult ADHD Rating Scale-Self-Report: Confirmation of Factor Structure in a Large Sample of Participants With ADHD

Objective: The Conners Adult ADHD Rating Scales (CAARS) assess symptoms specific to adults that are frequently used and have been translated into German. The current study tests the factor structure of the CAARS in a large sample of German adults with ADHD and compares the means of the CAARS subscales with those of healthy German controls. Method: CAARS were completed by 466 participants with ADHD and 851 healthy control participants. Confirmatory factor analysis was used to establish model fit with the American original. Comparisons between participants with ADHD and healthy controls and influences of gender, age, and degree of education were analyzed. Results: Confirmatory factor analysis showed a very good fit with the model for the American original. Differences between ADHD participants and healthy controls on all Conners Adult ADHD Rating Scales-Self-Report (CAARS-S) subscales were substantial and significant. Conclusion: The factor structure of the original American model was successfully replicated in this sample of adult German ADHD participants. (J. of Att. Dis. 2012; XX(X) 1-XX).

Large Sample Theory for Semiparametric Regression Models with Two-Phase, Outcome Dependent Sampling

Outcome-dependent, two-phase sampling designs can dramatically reduce the costs of observational studies by judicious selection of the most informative subjects for purposes of detailed covariate measurement. Here we derive asymptotic information bounds and the form of the efficient score and influence functions for the semiparametric regression models studied by Lawless, Kalbfleisch, and Wild (1999) under two-phase sampling designs. We show that the maximum likelihood estimators for both the parametric and nonparametric parts of the model are asymptotically normal and efficient. The efficient influence function for the parametric part aggress with the more general information bound calculations of Robins, Hsieh, and Newey (1995). By verifying the conditions of Murphy and Van der Vaart (2000) for a least favorable parametric submodel, we provide asymptotic justification for statistical inference based on profile likelihood.

Multinomial goodness-of-fit: large sample tests with survey design correction and exact tests for small samples

I introduce the new mgof command to compute distributional tests for discrete (categorical, multinomial) variables. The command supports largesample tests for complex survey designs and exact tests for small samples as well as classic large-sample x2-approximation tests based on Pearson’s X2, the likelihood ratio, or any other statistic from the power-divergence family (Cressie and Read, 1984, Journal of the Royal Statistical Society, Series B (Methodological) 46: 440–464). The complex survey correction is based on the approach by Rao and Scott (1981, Journal of the American Statistical Association 76: 221–230) and parallels the survey design correction used for independence tests in svy: tabulate. mgof computes the exact tests by using Monte Carlo methods or exhaustive enumeration. mgof also provides an exact one-sample Kolmogorov–Smirnov test for discrete data.

Self-esteem development across the life span: A longitudinal study with a large sample from Germany

The authors examined the development of self-esteem across the life span. Data came from a German longitudinal study with 3 assessments across 4 years of a sample of 2,509 individuals ages 14 to 89 years. The self-esteem measure used showed strong measurement invariance across assessments and birth cohorts. Latent growth curve analyses indicated that self-esteem follows a quadratic trajectory across the life span, increasing during adolescence, young adulthood, and middle adulthood, reaching a peak at age 60 years, and then declining in old age. No cohort effects on average levels of self-esteem or on the shape of the trajectory were found. Moreover, the trajectory did not differ across gender, level of education, or for individuals who had lived continuously in West versus East Germany (i.e., the 2 parts of Germany that had been separate states from 1949 to 1990). However, the results suggested that employment status, household income, and satisfaction in the domains of work, relationships, and health contribute to a more positive life span trajectory of self-esteem. The findings have significant implications, because they call attention to developmental stages in which individuals may be vulnerable because of low self-esteem (such as adolescence and old age) and to factors that predict successful versus problematic developmental trajectories.

Contextualizing local-scale point sample data using global-scale spatial datasets: Lessons learnt from the analysis of large-scale land acquisitions

This paper examines how the geospatial accuracy of samples and sample size influence conclusions from geospatial analyses. It does so using the example of a study investigating the global phenomenon of large-scale land acquisitions and the socio-ecological characteristics of the areas they target. First, we analysed land deal datasets of varying geospatial accuracy and varying sizes and compared the results in terms of land cover, population density, and two indicators for agricultural potential: yield gap and availability of uncultivated land that is suitable for rainfed agriculture. We found that an increase in geospatial accuracy led to a substantial and greater change in conclusions about the land cover types targeted than an increase in sample size, suggesting that using a sample of higher geospatial accuracy does more to improve results than using a larger sample. The same finding emerged for population density, yield gap, and the availability of uncultivated land suitable for rainfed agriculture. Furthermore, the statistical median proved to be more consistent than the mean when comparing the descriptive statistics for datasets of different geospatial accuracy. Second, we analysed effects of geospatial accuracy on estimations regarding the potential for advancing agricultural development in target contexts. Our results show that the target contexts of the majority of land deals in our sample whose geolocation is known with a high level of accuracy contain smaller amounts of suitable, but uncultivated land than regional- and national-scale averages suggest. Consequently, the more target contexts vary within a country, the more detailed the spatial scale of analysis has to be in order to draw meaningful conclusions about the phenomena under investigation. We therefore advise against using national-scale statistics to approximate or characterize phenomena that have a local-scale impact, particularly if key indicators vary widely within a country.

Correlation between dental care utilization and oral cavity cancer in a large sample of community-based United States residents

Cancer of the oral cavity and pharynx remains one of the ten leading causes of cancer death in the United States (US). Besides smoking and alcohol consumption, there are no well established risk factors. While poor dental care had been implicated, it is unknown if the lack of dental care, implying poor dental hygiene predisposes to oral cavity cancer. This study aimed to assess the relationship between dental care utilization during the past twelve months and the prevalence of oral cavity cancer. A cross-sectional design of the National Health Interview Survey of adult, non-institutionalized US residents (n=30,475) was used to assess the association between dental care utilization and self reported diagnosis of oral cavity cancer. Chi square statistic was used to examine the crude association between the predictor variable, dental care utilization and other covariates, while unconditional logistic regression was used to assess the relationship between oral cavity cancer and dental care utilization. There were statistically significant differences between those who utilized dental care during the past twelve months and those who did not with respect to education, income, age, marital status, and gender (p < 0.05), but not health insurance coverage (p = 0.53). Also, those who utilized dental care relative to those who did not were 65% less likely to present with oral cavity cancer, prevalence odds ratio (POR), 0.35, 95% Confidence Interval (CI), 0.12–0.98. Further, higher income advanced age, people of African heritage, and unmarried status were statistically significantly associated with oral cavity cancer, (p < 0.05), but health insurance coverage, alcohol use and smoking were not, p > 0.05. However, after simultaneously controlling for the relevant covariates, the association between dental care and oral cavity cancer did not attenuate nor persist. Thus, compared with those who did not use dental care, those who did wee 62% less likely to present with oral cavity cancer adjusted POR, 0.38, 95% CI, 0.13-1.10. Among US adults residing in community settings, use of dental care during the past twelve months did not significantly reduce the predisposition to oral cavity cancer. However, due to the nature of the data used in this study, which restricts temporal sequence, a large sample prospective study that may identify modifiable factors associated with oral cancer development namely poor dental care, is needed. ^

Multinomial goodness-of-fit: large sample tests with survey design correction and exact tests for small samples

A new Stata command called -mgof- is introduced. The command is used to compute distributional tests for discrete (categorical, multinomial) variables. Apart from classic large sample $\chi^2$-approximation tests based on Pearson's $X^2$, the likelihood ratio, or any other statistic from the power-divergence family (Cressie and Read 1984), large sample tests for complex survey designs and exact tests for small samples are supported. The complex survey correction is based on the approach by Rao and Scott (1981) and parallels the survey design correction used for independence tests in -svy:tabulate-. The exact tests are computed using Monte Carlo methods or exhaustive enumeration. An exact Kolmogorov-Smirnov test for discrete data is also provided.

Variance components analyses of multiple asthma traits in a large sample of Australian families ascertained through a twin proband

Background: Intermediate phenotypes are often measured as a proxy for asthma. It is largely unclear to what extent the same set of environmental or genetic factors regulate these traits. Objective: Estimate the environmental and genetic correlations between self-reported and clinical asthma traits. Methods: A total of 3073 subjects from 802 families were ascertained through a twin proband. Traits measured included self-reported asthma, airway histamine responsiveness (AHR), skin prick response to common allergens including house dust mite (Dermatophagoides pteronyssinus [D. pter]), baseline lung function, total serum immunoglobulin E (IgE) and eosinophilia. Bivariate and multivariate analyses of eight traits were performed with adjustment for ascertainment and significant covariates. Results: Overall 2716 participants completed an asthma questionnaire and 2087 were clinically tested, including 1289 self-reported asthmatics (92% previously diagnosed by a doctor). Asthma, AHR, markers of allergic sensitization and eosinophilia had significant environmental correlations with each other (range: 0.23-0.89). Baseline forced expiratory volume in 1 s (FEV1) showed low environmental correlations with most traits. Fewer genetic correlations were significantly different from zero. Phenotypes with greatest genetic similarity were asthma and atopy (0.46), IgE and eosinophilia (0.44), AHR and D. pter (0.43) and AHR and airway obstruction (-0.43). Traits with greatest genetic dissimilarity were FEV1 and atopy (0.05), airway obstruction and IgE (0.07) and FEV1 and D. pter (0.11). Conclusion: These results suggest that the same set of environmental factors regulates the variation of many asthma traits. In addition, although most traits are regulated to great extent by specific genetic factors, there is still some degree of genetic overlap that could be exploited by multivariate linkage approaches.