What Causes the Mean Bias of the Likelihood Ratio Statistic with Many Variables?

Abstract

Survey data often contain many variables. Structural equation modeling (SEM) is commonly used in analyzing such data. However, conventional SEM methods are not crafted to handle data with a large number of variables (p). A large p can cause T_ml, the most widely used likelihood ratio statistic, to depart drastically from the assumed chi-square distribution even with normally distributed data and a relatively large sample size N. A key element affecting this behavior of T_ml is its mean bias. The focus of this article is to determine the cause of the bias. To this end, empirical means of T_ml via Monte Carlo simulation are used to obtain the empirical bias. The most effective predictors of the mean bias are subsequently identified and their predictive utility examined. The results are further used to predict type I errors of T_ml. The article also illustrates how to use the obtained results to determine the required sample size for T_ml to behave reasonably well. A real data example is presented to show the effect of the mean bias on model inference as well as how to correct the bias in practice.     

BIS, BAS, and bias: The role of personality and cognitive bias in social anxiety

The goal of the present research was to test the hypothesis that cognitive biases for negative and threatening social information mediate the effects of behavioral inhibition system (BIS) and behavioral approach system (BAS) sensitivity on social anxiety. Participants completed self-report measures of BIS and BAS and then underwent a social-threat induction procedure in which they were told they would have to perform a speech. A battery of cognitive bias measures was then administered, followed by a battery of state anxiety measures. Audience members also rated participants’ anxiety during the speech. Structural equation modeling was used to test the hypothesized model. As predicted, the fully-mediated model showed the best fit to the data, and higher BIS and lower BAS were found to have significant indirect effects on social anxiety via cognitive bias.     

Measuring and Reducing Religious Bias in Post‐Conflict Zones: Evidence from Côte d'Ivoire

This article explores how religious bias, once it has been cultivated through politicization and violence, can be reduced. Using foundations from social identity theory and superordinate goal theory, I develop post‐conflict bias reduction strategies that include competing types of superordinate messages, economic and theological, as well as different sources of those messages. To test these strategies, I use video‐based information treatments coupled with Implicit Association Tests in Bouaké, Côte d'Ivoire. The experimental findings point to three conclusions. First, implicit Muslim‐Christian bias in the study area remains high. Second, Christians in the study tend to be more biased against Muslims than Muslims are against Christians. Third, the effectiveness of treatments depends on the subjects who receive those treatments: theological messages are most effective in reducing bias among Muslims, regardless of their source, and strategies that rely on political leaders to deliver messages perform best among Christians, regardless of the content.     

The detection and influence of problematic item content in ability tests: An examination of sensitivity review practices for personnel selection test development

In organizational and educational practices, sensitivity reviews are commonly advocated techniques for reducing test bias and enhancing fairness. In the present paper, results from two studies are reported which investigate how effective individuals are at detecting problematic test content and the influence such content has on important testing outcomes. In Study 1, signal detection analyses are used to examine the role of individual differences in the identification of insensitive test items, while Study 2 investigates the extent to which insensitivity differentially influences item performance and reactions. Results revealed small but significant differences in the overall accuracy and response tendencies of student test reviewers on the basis of demographics and key individual differences variables. Contrary to predictions however, problematic items did not exhibit differential item functioning across sex nor did their presence engender negative test taker reactions. Implications and suggestions for future research and sensitivity review practices are discussed.     

Eye-tracking the own-race bias in face recognition: Revealing the perceptual and socio-cognitive mechanisms

Own-race faces are recognised more accurately than other-race faces and may even be viewed differently as measured by an eye-tracker (Goldinger, Papesh, & He, 2009). Alternatively, observer race might direct eye-movements (Blais, Jack, Scheepers, Fiset, & Caldara, 2008). Observer differences in eye-movements are likely to be based on experience of the physiognomic characteristics that are differentially discriminating for Black and White faces. Two experiments are reported that employed standard old/new recognition paradigms in which Black and White observers viewed Black and White faces with their eye-movements recorded. Experiment 1 showed that there were observer race differences in terms of the features scanned but observers employed the same strategy across different types of faces. Experiment 2 demonstrated that other-race faces could be recognised more accurately if participants had their first fixation directed to more diagnostic features using fixation crosses. These results are entirely consistent with those presented by Blais et al. (2008) and with the perceptual interpretation that the own-race bias is due to inappropriate attention allocated to the facial features ( and ).