Search

Copyright of BMS: Bulletin de Methodologie Sociologique (Sage Publications Ltd.) is the property of Sage Publications Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

We examine several approaches for inferring logit models from empirical margins of predictor covariates and conditional margins containing the means of a binary response for each covariate margin. One method is to fit proxy data to the conditional response using the beta distribution, a process we call 'margin analysis.' Proxy data can obtained using three approaches: (1) implementing the iterative proportional fitting (IPF) procedure on the margin totals, (2) sampling from a larger relevant data source such as the census, and (3) enumerating, or sampling from, the combinatoric space of all possible tables constrained by the margins. The first procedure is a well-studied approach for estimating contingency tables from margins, but it does not necessarily maintain the associations between the covariates unless seeded with an initial table containing those associations. In the second approach, which is appropriate for analyzing sociodemographic covariates, we can use a large census sample adjusting for sampling biases observed in the empirical margins. However, the appropriateness of using a census proxy depends substantially on how similar the sampling pools are. Our third approach entails exploring the combinatoric space of all contingency tables constrained by the margins while considering the associations among the covariates. We aggregate the logit models estimated from each table in that space into a single model. This approach is more robust than the first two as it considers multiple proxies. While the estimated logit models from each approach are generally similar to one another, for the low-dimensional tables we explore in this paper, the combinatoric approach incurs wider standard errors, which renders potentially significant coefficients insignificant. Finally, we suggest weighting the combinatoric models with evidence-relevant probabilities obtained using the multivariate Pólya distribution. [ABSTRACT FROM AUTHOR]

The author offers a response to the articles "Snowball Sampling," published in a 1961 issue of "Annals of Mathematical Statistics," and "Respondent-Driven Sampling: An Assessment of Current Methodology," published in a 2010 issue of the publication. It is stated that snowball sampling in populations difficult to reach for research purposes is not the same as response-driven sampling but, the two methods can be combined for conducting research on such groups.

Many basic questions in the social network literature center on the distribution of aggregate structural properties within and across populations of networks. Such questions are of increasing relevance given the growing availability of network data suitable for meta-analytic studies, as well as the rise of study designs that involve the collection of data on multiple networks drawn from a larger population. Despite this, little work has been done on model-based inference for the properties of graph populations, or on methods for comparing such populations. Here, we attempt to rectify this gap by introducing a family of techniques that combines an existing approach to the identification of structural biases in network data (the use of conditional uniform graph quantiles) with strategies drawn from nonparametric Bayesian analysis. Conditional uniform graph quantiles are the quantiles of an observed structural property in the reference distribution produced by evaluating that property over all graphs with certain fixed characteristics (e.g., size or density). These quantiles have long been used to measure the extent to which a property of interest on a single network deviates from what would be expected given that network's other characteristics. The methods introduced here employ such quantile information to allow for principled inference regarding the distribution of structural biases within (and comparison across) populations of networks, given data sampled at the network level. The data requirements of these methods are minimal, thus making them well-suited to meta-analytic applications for which complete network data (as opposed to summary statistics) are often unavailable. The structural biases inferred using these methods can be expressed in terms of posterior predictives for familiar and easily communicated quantities, such as p-values. In addition to the methods themselves, we present algorithms for posterior simulation from this model class, illustrating their use with applications to the analysis of social structure within urban communes and radio communications among emergency personnel. We also discuss how this approach may applied to quantiles arising from other reference distributions, such as those obtained using general exponential-family random graph models. [ABSTRACT FROM AUTHOR]

The author offers opinions on the article "On Respondent-Driven Sampling and Snowball Sampling in Hard to Reach Populations and Snowball Sampling not in Hard-to-Reach Populations," by Leo A. Goodman, published in this issue. It is stated that confusion between snowball sampling as a methodology in statistics and respondent-driven sampling RDS) stems from multiple forms of the estimator in the RDS process stemming from the incremental development of the methodology.

Recent research has shown that two entropy-based segregation indices possess an appealing mixture of basic and subsidiary but useful properties. It would appear that the only fundamental difference between the mutual information or M index, and the entropy information or H index, is that the second is a normalized version of the first. This paper introduces another normalized index in that family, the H* index, which captures segregation as the tendency of racial groups to have different distributions across schools. More importantly, the paper shows that applied researchers may do better using the M index than using either H or H* in two circumstances: (1) if they are interested in the decomposability of the measurement of segregation, and (2) if they are interested in a margin-free measurement of segregation changes. The shortcomings of the H and H* indices are illustrated below by means of numerical examples, as well as with school segregation data by ethnic group in the U.S. public school system between 1989 and 2005. [ABSTRACT FROM AUTHOR]

An introductions is presented to articles published in this issue including "How Not to Lie With Ethnography," "Inferring Logit Models from Emplrical Margins Using Proxy Data" and "Entropy-Based Segregation Indices."

The author offers opinions on the article "On Respondent-Driven Sampling and Snowball Sampling in Hard to Reach Populations and Snowball Sampling not in Hard-to-Reach Populations," by Leo A. Goodman, published in this issue. It is stated that the term snowball sampling has come to refer to two different means of conducting sampling in sociology research, obscuring the use of respondent-driven sampling which is seen as superior methodology when studying hard to reach populations.

General random graphs (i.e., stochastic models for networks incorporating heterogeneity and/or dependence among edges) are increasingly in wide use in the study of social and other networks, but few techniques other than simulation have been available for studying their behavior. On the other hand, random graphs with independent edges (i.e., the Bernoulli graphs) are well-studied, and a large literature exists regarding their properties. In this paper, we demonstrate a method for leveraging this knowledge by constructing families of Bernoulli graphs that bound the behavior of an arbitrary random graph in a well-defined sense. By studying the behavior of these Bernoulli graph bounds, we can thus constrain the properties of a given random graph. We illustrate the utility of this approach via application to several problems from the social network literature, including identifying degeneracy in Markov graph models, studying the potential impact of tie formation mechanisms on epidemic potential in sexual contact networks, and robustness testing of inhomogeneous Bernoulli models based on geographical covariates. Practical heuristics for assessing bound tightness and guidance for use in theoretical and methodological applications are also discussed. [ABSTRACT FROM AUTHOR]

The theoretical consequences of measurement error in outcome variables that are continuous are widely known by practitioners, at least for the classical model: purely random errors will lead to a loss of efficiency but not to bias in regression coefficients. When the outcome variable is binary, however, regression coefficients, both linear and nonlinear, will contain bias, even if the measurement error (in this setting more commonly referred to as classification error) is purely random. This paper illustrates a method of correcting for misclassification bias that relies solely on the primary survey data. It is particularly suited to analyses of surveys where external validation of survey responses is unavailable but where there is strong reason to suspect contaminated data. This situation is common in observational studies of the health of populations. The technique is applied to a model of the antecedents of post-traumatic stress disorder (PTSD) using data from a large-scale cross-sectional survey of Vietnam-era veterans. Results show that when adjusted for errors in diagnoses, the sample PTSD prevalence estimate falls significantly; that failure to correct for misclassification in PTSD dramatically understates the effects of risk factors; and that this downward bias remains even when the model incorporates differential classification errors-that is, errors that are correlated with some of the explanatory variables in the model. [ABSTRACT FROM AUTHOR]

The article describes an alternative approach of gathering in-depth qualitative accounts and content coding them for statistical analysis, wherein the regression coefficients of workplace ethnographies were introduced into a body of sociological research. It is attested that the merging of qualitative and quantitative approaches in sociology is advancing through methods such as appending comprehensive interviews to surveys. It is further posited that funding requirements for such projects are minimal and students can easily be involved. The authors conclude by discussing other bodies of ethnographic research that are ready for systematic analysis.

Respondent-driven sampling (RDS) employs a variant of a link-tracing network sampling strategy to collect data from hard-to-reach populations. By tracing the links in the underlying social network, the process exploits the social structure to expand the sample and reduce its dependence on the initial (convenience) sample. The current estimators of population averages make strong assumptions in order to treat the data as a probability sample. We evaluate three critical sensitivities of the estimators: (1) to bias induced by the initial sample, (2) to uncontrollable features of respondent behavior, and (3) to the without-replacement structure of sampling. Our analysis indicates: (1) that the convenience sample of seeds can induce bias, and the number of sample waves typically used in RDS is likely insufficient for the type of nodal mixing required to obtain the reputed asymptotic unbiasedness; (2) that preferential referral behavior by respondents leads to bias; (3) that when a substantial fraction of the target population is sampled the current estimators can have substantial bias. This paper sounds a cautionary note for the users of RDS. While current RDS methodology is powerful and clever, the favorable statistical properties claimed for the current estimates are shown to be heavily dependent on often unrealistic assumptions. We recommend ways to improve the methodology. [ABSTRACT FROM AUTHOR]

This paper proposes a two-stage maximum likelihood (ML) approach to normal mixture structural equation modeling (SEM) and develops a statistical inference that allows distributional misspecification. Saturated means and covariances are estimated at stage 1 together with a sandwich-type covariance matrix. These are used to evaluate structural models at stage 2. Techniques accumulated in the conventional SEM literature for model diagnosis and evaluation can be used to study the model structure for each component. Examples show that the two-stage ML approach leads to correct or nearly correct models even when the normal mixture assumptions are violated and initial models are misspecified. Compared to single-stage ML, two-stage ML avoids the confounding effect of model specification and the number of components, and it is computationally more efficient. Monte Carlo results indicate that two-stage ML loses only minimal efficiency under the condition where single-stage ML performs best. Monte Carlo results also indicate that the commonly used model selection criterion BIC is more robust to distribution violations for the saturated model than that for a structural model at moderate sample sizes. The proposed two-stage ML approach is also extremely flexible in modeling different components with different models. Potential new developments in the mixture modeling literature can be easily adapted to study issues with normal mixture SEM. [ABSTRACT FROM AUTHOR]

Arguably, sociologists often apply deterministic explanations, unlike in many other sciences where probabilistic explanations have turned out to be more efficient. The advantages of applying probabilistic explanations more frequently could be: improved possibilities for cross-disciplinary work; the simplification of successful explanations; an easier way to implement uncertainty to models; new ways to utilize the classical theories of sociology; and increased freedom to choose research methods. According to the previously published literature, one of the reasons for not applying probabilistic explanations has been the strong positivist heritage in sociology. The following additional factors are considered: important classics of sociology were written before probability theories matured; and qualitative methods have provided the possibility to respond to the need for empirical research. In this way, sociologists have avoided the extensive work of re-evaluating the existing theories from a probabilistic point of view. [ABSTRACT FROM AUTHOR]

The article presents a research response to the methodological criticism issued by Tom Van der Meer, Manfred Te Grotenhuis, and Ben Pelzer included within the issue, challenging the case examples provided in the author's sociological analysis of religiosity and volunteering behavior in the United States. The authors disagree with the conclusions made against their original study but also compliment the methodology suggestions raised by the article and provide alternative paradigms for future studies to correct for the problems.

This study assessed agreement level about the occurrence of past-year male-to-female partner violence (MFPV) and female-to-male partner violence (FMPV) among a sample of 897 blue-collar couples. Intimate partner violence (IPV) was measured with the Physical Assault subscale of the revised Conflict Tactics Scales (CTS2). Agreement level was assessed with Cohen's kappa statistic. Lower-bound estimates (based on couple agreement that an IPV event occurred) and upper-bound estimates (based on uncorroborated reports from either partner that an IPV event occurred) were calculated. Results indicated low agreement for most IPV behaviors (κ < .40). Estimated lower- and upper-bound rates for MFPV were 6.7% and 21.2%, for FMPV, 7.1% and 24.2%, and for any IPV, 10.1% and 30.2%. Findings suggest that single-point IPV prevalence estimates are biased; lower- and upper-bound estimates using collateral reports should be calculated when possible. In addition, findings underscore the importance of conducting IPV research among understudied populations, such as working-class couples, that may be at elevated IPV risk. [ABSTRACT FROM AUTHOR]

This paper examines the development of SPSS from 1968 to 2008, and the manner in which it has been used in teaching and research in British Sociology. We do this in order to reveal some of the changes that have taken place in statistical reasoning as an inscription device in the discipline over this period. We conclude that to characterise these changes as a shift from ‘causal’ to more ‘descriptive’ modes of analysis is too simplistic. Such a shift is certainly apparent, but it meshes in complex ways with a range of other – just as important – changes, that together mark a phase-shift in the functioning of sociological quantification. [ABSTRACT FROM AUTHOR]

The article discusses various reports published within the issue, including one by Uprichard, Burrows, and Byrne on the history of the statistical software package SPSS (the Statistical Package for the Social Sciences), one by Savage about the deployment of in-depth interviews, and another one by Stanley on the diaries and directives of Mass-Observation.

Data editing, a crucial task in the data production process, has received little scientific attention. Consequently, there is no consensus among social scientists about how data should be edited or by whom. While some argue that it should be left to data managers and data users, others claim that it is primarily a task for fieldworkers. The authors review these divergent approaches to editing and evaluate the underlying theoretical arguments. Results are reported from a methodological experiment in which different types of actors who are party to the data production and research process were asked to solve artificially generated inconsistencies in real survey data. Results are informative on two counts. First, the least accurate editors were the researchers with no field experience in the survey sites. Second, when provided with only partial information on which in make editing decisions, fieldworkers edited more accurately than both data managers and data users. [ABSTRACT FROM AUTHOR]

This paper describes and contrasts two useful ways to employ a latent class variable as a mixture variable in regression analyses of panel data with a categorical dependent variable. One way is to model unobserved heterogeneity in the trajectory, or change in the distribution, of the dependent variable. Two models that accomplish this are the latent trajectory model and latent growth curve model for a categorical dependent variable having ordered categories. Each latent class here represents a distinct trajectory of the dependent variable. The latent trajectory model introduces covariate effects on the composition of latent classes, while the latent growth curve model introduces covariate effects on both the “intercept” and the “slope” of growth in logit, which may vary among latent classes. The other useful way is to model unobserved heterogeneity in the state dependence of the dependent variable. Two models that accomplish this are introduced for a simultaneous analysis of response probability and response stability, and the latent class variable is employed to distinguish two latent populations that differ in the stability of responses over time. One of them is the switching multinomial logit model with a time-lagged dependent variable as its separation indicator, and the other is the mover-stayer regression model. By applying these four models to empirical data, this paper demonstrates the usefulness of these models for panel-data analyses. Example programs for specifying these models based on the LEM program are also provided. [ABSTRACT FROM AUTHOR]

This article questions whether implication analysis, an approach to statistical sociology described by Stanley Lieberson and Joel Horwich elsewhere in this journal, is a novel approach to sociology. The author reframes Lieberson and Horwich's criticism of statistical methods in sociology as a condemnation of scholarly research based on weak data. Questions are also raised about the use of science as an example for sociology. The author suggests that an investigation of the relationship between experimental psychology and predictive approaches to sociology could improve the credibility of that field.

This article discusses the questions about experimental design raised by Stanley Lieberson and Joel Horwich elsewhere in this issue. Lieberson and Horwich advanced the theory of implication analysis as an alternative to experimental designs in sociology that prioritize falsifiability. The potential contribution of implication analysis to the design of statistical regression models in sociological research is described. Questions are raised about how the fairness of the Lieberson and Horwich's analysis of scientific thought as it relates to sociological analysis. Implication analysis is considered in terms how its application would change the emphasis given to control variables and alternative theories in the development of sociological research.

This article responds to an article on the design of sociological experiments written by Stanley Lieberson and Joel Horwich and published elsewhere in this journal. The weighting of evidence in the testing of sociological hypotheses is addressed. Dichotomies between induction and deduction in the analysis of sociological data are criticized in favor of abductive methods in sociological research. The design of an experiment intended to determine whether chemical and biological weapons were used in conflicts in Laos, Cambodia, and Afghanistan during the 1970s is considered.

This article responds to the development of implication analysis, an approach to sociological research that emphasizes thorough analysis of the theoretical context within which data are analysed. This approach is developed in an essay by Joel Horwich and Stanley Lieberson elsewhere in this issue. The application of Horwich and Lieberson's approach to the construction of doctoral dissertations is framed in terms of the limitations that it would place on the creativity of students. Issues relating to the generation of comparative data to carry out the projects of implication analysis are considered.

The increase in life expectancy at middle age has added quality of life to the policy concerns related to older ages. The present article asks whether this quality of life varies with socio-economic position. Using data from the English Longitudinal Study of Ageing (n.d.), the article answers this question in terms of the National Statistics Socio-Economic Classification and CASP-19, a measure of positive quality of life in early old age. Among all respondents aged 50 to 75 years, quality of life was found to be graded by social position, with the difference in quality of life between the higher managerial and professional group and the routine group being of comparable size to having a limiting long-standing illness; with the size of this difference varying between labour market sub-groups. The implication of these findings for ageing policy and for the National Statistics Socio-Economic Classification are discussed. [ABSTRACT FROM AUTHOR]

This article examines more carefully the oft-made hypotheses that (1) ‘technocrats’ or politicians with an economics background are increasingly common and (2) that this ‘improvement’ in qualifications will lead to improvements in economic policy. The article presents a database on the qualifications of leaders of the world's major countries over the past four decades. The article finds that while there is evidence for increasing ‘technification,’ there are also distinct and persistent historical patterns among Asian, African, Middle Eastern, and Latin American leaders. Using statistical analysis, the article finds that we cannot conclude that leadership training in economics leads to better economic outcomes. [ABSTRACT FROM AUTHOR]

This article is a follow-up to Jones, Nagin, and Roeder (2001), which described an SAS procedure for estimating group-based trajectory models. Group-based trajectory is a specialized application of finite mixture modeling and is designed to identify clusters of individuals following similar progressions of some behavior or outcome over age or time. This article has two purposes. One is to summarize extensions of the methodology and of the SAS procedure that have been developed since Jones et al. The other is to illustrate how group-based trajectory modeling lends itself to presentation of findings in the form of easily understood graphical and tabular data summaries. [ABSTRACT FROM AUTHOR]

Surveys often sample adults across a broad range of ages, measuring the same outcomes in several interviews spaced during a period of years and comparing the changes observed across segments of the adult life course. Put in sequence, those change vectors provide a composite image of the outcome's life course trajectory. To illustrate, the authors estimate depression vectors in a sample of U.S. adults ages 18 and older at baseline in 1995, with follow-up interviews in 1998 and 2001. They show the vector equations and their graphs and also their synthetic-cohort projection. The authors introduce the trend-function and virtual-cohort projection, showing how they provide tests of "convergence" and other hypotheses about trajectories and trends. Results show depression dropping and then rising across adulthood more steeply than suggested by cross-sectional differences among age groups. They also indicate a rise and fall in age-specific levels of depression across cohorts. [ABSTRACT FROM AUTHOR]

McDonald and Moffitt's (1980) decomposition method for Tobit coefficients disaggregates the total changes in a censored outcome variable into the changes in the probability of attaining noncensored values and the changes in the noncensored values themselves. McDonald and Moffitt, however, did not highlight two limitations of this decomposition, and sociological applications of the Tobit decomposition that do not acknowledge these limitations have been misleading. First, the decomposition ratio of the two effects depends on the choice of an observation point in the sample. Some sociological studies invalidly generalized a decomposition at the mean point of the sample to the whole sample. Second, once the observation point is fixed, the decomposition ratio is the same across all coefficients or across all predictors. That is, the two effects remain proportionally constant, but studies often erroneously assume they differ across predictors. Further implications are discussed in relation to other selection bias models. [ABSTRACT FROM AUTHOR]

As the counterfactual model of causality has increased in popularity, sociologists have returned to matching as a research methodology. In this article, advances over the past two decades in matching estimators are explained, and the practical limitations of matching techniques are emphasized. The authors introduce matching methods by focusing first on ideal scenarios in which stratification and weighting procedures warrant causal inference. Then, they discuss how matching is often undertaken in practice, offering an overview of the most prominent data analysis routines. With four hypothetical examples, they demonstrate how the assumptions behind matching estimators often break down in practice. Even so, the authors argue that matching techniques can be used effectively to strengthen the prosecution of causal questions in sociology. [ABSTRACT FROM AUTHOR]

The authors compare the operating characteristics of the bootstrap-F approach, a direct extension of the work of Berkovits, Hancock, and Nevitt, with Huynh's improved general approximation (IGA) and the Brown-Forsythe (BF) multivariate approach in a mixed repeated measures design when normality and multisample sphericity assumptions do not hold. The results of the simulation show that the three approaches adequately control Type I error when data are generated from normal or slightly nonnormal distributions. However, when data are generated from distributions with moderate or severe skewness, the approaches tend to produce conservative Type I error rates, except the IGA test of the main effect, which has liberal Type I error rates in some conditions. With regard to power, it was found that the bootstrap-F approach can compete with the IGA approach but not with the BF approach: The power differences favoring the bootstrap-F approach are generally small, whereas those favoring the BF approach are substantial. [ABSTRACT FROM AUTHOR]

The most promising class of statistical models for expressing structural properties of social networks observed at one moment in time is the class of exponential random graph models (ERGMs), also known as p* models. The strong point of these models is that they can represent a variety of structural tendencies, such as transitivity, that define complicated dependence patterns not easily modeled by more basic probability models. Recently, Markov chain Monte Carlo (MCMC) algorithms have been developed that produce approximate maximum likelihood estimators. Applying these models in their traditional specification to observed network data often has led to problems, however, which can be traced back to the fact that important parts of the parameter space correspond to nearly degenerate distributions, which may lead to convergence problems of estimation algorithms, and a poor fit to empirical data. This paper proposes new specifications of exponential random graph models. These specifications represent structural properties such as transitivity and heterogeneity of degrees by more complicated graph statistics than the traditional star and triangle counts. Three kinds of statistics are proposed: geometrically weighted degree distributions, alternating k-triangles, and alternating independent two-paths. Examples are presented both of modeling graphs and digraphs, in which the new specifications lead to much better results than the earlier existing specifications of the ERGM. It is concluded that the new specifications increase the range and applicability of the ERGM as a tool for the statistical analysis of social networks. INSETS: Figure 1: Some configurations for nondirected graphs.;Figure 2: Partial conditional dependence when four-cycle...;Figure 3: Two examples of a two-triangle.;Figure 4: A k-triangle for k=5; also called a five-triangle [ABSTRACT FROM AUTHOR]

The article presents a teaching technique designed to motivate students to learn statistics. The author gives students a list of common occupations for people with a degree in Sociology, their salaries, and the use of statistics in the occupation. The goals of this technique is to convince students that statistics is a worthwhile subject to learn, to reduce student anxiety by letting them demonstrate that they can do statistics, and to provide an example that can be used throughout the semester to illustrate multiple statistical techniques. The approach simply acknowledges that money matters to students. It uses the promise of monetary reward to give them at least one reason to pay attention and learn throughout the rest of the course.

Five experiments document biases in the way people predict the outcomes of interdependent social situations. Participants predicted that situational constraints would restrain their own behavior more than it would the behavior of others, even in situations where everyone faced identical constraints. When anticipating the effects of deadlines on outcomes of negotiations, participants predicted that deadlines would hinder their performance more than it would hinder the performance of others. The results shed light on the psychological processes by which people predict the outcomes of and select strategies in strategic social interaction. They extend prior findings, such as people believing themselves to be below average on difficult tasks, to highly interdependent situations. Furthermore, the article shows both how focusing can account for these effects and also how perspective taking can reduce their biasing influence. [ABSTRACT FROM AUTHOR]

This article discusses the use of concrete historical examples to enhance learning in statistics classes. A host of scholars have acknowledged the difficulties experienced by students learning research methods and statistics, especially those in the social sciences who may not be as mathematically inclined as students in other fields of study. Students' anxieties may cause them to delay taking methods courses, adversely affecting their ability to understand material in other classes that incorporate these tools. Many suggestions for easing student anxiety have been offered, including the use of humor, small group work, and use of local county data. Some students never seemed to get over their anxiety about statistics. Students occasionally returned a year or two later to ask questions that indicated that they had not retained, at least with any confidence, the material they had learned in the classroom. Two approaches were developed to deal with these concerns of learning and retention. First, instead of reliance upon lecture notes per se, the plan of instruction instead incorporated data analysis learning objectives. The second instructional approach included finding and applying examples that would meet several criteria for sustaining student interest and enhancing their learning and confidence with statistics.

In this article, the authors examine the most common type of improper solutions: zero or negative error variances. They address the causes of, consequences of, and strategies to handle these issues. Several hypotheses are evaluated using Monte Carlo simulation models, including two structural equation models with several misspecifications of each model. Results suggested several unique findings. First, increasing numbers of omitted paths in the measurement model were associated with decreasing numbers of improper solutions. Second, bias in the parameter estimates was higher in samples with improper solutions than in samples including only proper solutions. Third, investigation of the consequences of using constrained estimates in the presence of improper solutions indicated that inequality constraints helped some samples achieve convergence. Finally, the use of confidence intervals as well as four other proposed tests yielded similar results when testing whether the error variance was greater than or equal to zero. [ABSTRACT FROM AUTHOR]

Statistical methods have had a successful half-century in sociology, contributing to a greatly improved standard of scientific rigor in the discipline. I identify three overlapping postwar generations of statistical methods in sociology, based on the kinds of data they address. The first generation, which started in the late 1940s, deals with cross-tabulations and focuses on measures of association and log-linear models, perhaps the area of statistics to which sociology has contributed the most. The second generation, which began in the 1960s, deals with unit-level survey data and focuses on LISREL-type causal models and event-history analysis. The third generation, starting to emerge in the late 1980s, deals with data that do not fall easily into either of these categories, either because they have a different form, such as texts or narratives, or because dependence is a crucial aspect, as with spatial or social network data. There are many new challenges, and the area is ripe for statistical research; several major institutions have recently launched new initiatives in statistics and the social sciences. [ABSTRACT FROM AUTHOR]

This article presents the author's comments on a work of Steven N. Durlauf on the potential applications and extensions for a binary choice-based social interaction framework. Grounded in economic theory, this framework formalizes the influences of social structure in decision processes through individual preferences, which embody feelings about the choice in isolation and in relation to what is known about the choices made by others. The statistical model is embedded in the theoretical framework and as such adds much needed analytic clarity. The framework has several useful features that allow for a description of the whole system, not just individual behavior. Its representation of what social influences are may be simplistic but that is a tradeoff for greater generality that makes the framework applicable to diverse areas and broad theories, and more importantly, permits it to encompass ostensibly dissimilar processes and structures. The framework provides the foundation from which others can elaborate and add specificity, and it provides the basis for theoretically informed empirical analyses that bear on longstanding debates in the social sciences and have practical significance for public policy.

A categorical profile is a vector of observed values of several categorical variables that share a common context. Statistical analysis of categorical profiles may involve study of the joint distribution of the profiles or study of the relationship of the profiles to explanatory variables. Such analysis entails special difficulties due to the very large number of possible categorical profiles and due to the very strong relationships among the responses. Given the complex nature of the relationships among the variables, large samples are required for analysis. These large samples generally render useless traditional methods of model fitting based on tests of goodness of fit. Alternatively, model quality may be assessed in terms of descriptive power, as measured by information-theoretic criteria. A useful method of data description involves summary statistics derived from the data. A new approach combining log-linear models and summary statistics results in insightful and parsimonious description of categorical profiles. The analysis of summary statistics results in the use of log-linear models that differ substantially from those commonly employed in the analysis of profile data. Special measures are introduced for comparison of the descriptive power associated with different choices of summary statistics and for comparison of the number of parameters required for each model. Estimates for these special measures are proposed, and large-sample properties are considered in order to find asymptotic confidence intervals, providing an added inferential value to the proposed methods of analysis. Through use of an empirical example of responses to questions concerning legal abortions, it is shown that models based on very succinct summaries of responses involve remarkably little information loss, thus describing the data relatively accurately and parsimoniously. [ABSTRACT FROM AUTHOR]

We introduce some simple models that can be viewed as special cases of the general model and general approach to analyzing the association in multiway tables that we first presented in 1998. We use two empirical examples to demonstrate the flexibility and utility of the simple models introduced here and of our general approach. [ABSTRACT FROM AUTHOR]

A class of statistical models is proposed for longitudinal network data. The dependent variable is the changing (or evolving) relation network, represented by two or more observations of a directed graph with a fixed set of actors. The network evolution is modeled as the consequence of the actors making new choices, or withdrawing existing choices, on the basis of functions, with fixed and random components, that the actors try to maximize. Individual and dyadic exogenous variables can be used as covariates. The change in the network is modeled as the stochastic result of network effects (reciprocity, transitivity, etc.) and these covariates. The existing network structure is a dynamic constraint for the evolution of the structure itself. The models are continuous-time Markov chain models that can be implemented as simulation models. The model parameters are estimated from observed data. For estimating and testing these models, statistical procedures are proposed that are based on the method of moments. The statistical procedures are implemented using a stochastic approximation algorithm based on computer simulations of the network evolution process. [ABSTRACT FROM AUTHOR]

We propose an alternative method of conducting exploratory latent class analysis that utilizes latent class factor models, and compare it to the more traditional approach based on latent class cluster models. We show that when formulated in terms of R mutually independent, dichotomous latent factors, the LC factor model has the same number of distinct parameters as an LC cluster model with R+1 clusters. Analyses over several data sets suggest that LC factor models typically fit data better and provide results that are easier to interpret than the corresponding LC cluster models. We also introduce a new graphical “bi-plot” display for LC factor models and compare it to similar plots used in correspondence analysis and to a barycentric coordinate display for LC cluster models. New results on identification of LC models are also presented. We conclude by describing various model extensions and an approach for eliminating boundary solutions in identified and unidentified LC models, which we have implemented in a new computer program. [ABSTRACT FROM AUTHOR]

In this article, the author discusses on contradictions between generality and accuracy issues in sociological methodologies with considering the critical article by David L. Weakliem about the Bayesian information critenon (BIC). In sampling, efficient samples are often expensive to collect. In estimation, robust estimators tend to be inefficient. In data analysis, parsimonious models may not fit observed data. These kind of so-called trade-offs create tensions because researchers ideally prefer efficiency and low cost in sampling, robustness and efficiency in estimation, and parsimony and good fit in data analysis. Weakliem has pointed out two major limitations of the BIC. At first, Weakliem shows that Bayes factors depend on priors pertaining to the distribution of unknown parameters, whereas the BIC does not. Second, the BIC takes into consideration only the total sample size but not its distributions across tables, categories, or cells. Weakliem makes a good case that the BIC has certain limitations. The key question is whether the limitations are so severe as to render the BIC useless in applied research.

This article presents a commentary evaluation of the critical article, by David L. Weakliem, on the Bayesian Information Criterion (BIC) for Model Selection, contained in this issue of the journal "Sociological Methods & Research." The author describes that Weakliem makes two main points about the BIC approximation. The first is that BIC yields an approximation to Bayes factors that corresponds closely to a particular prior on the model parameters. This may or may not be similar to the researcher's actual prior information or beliefs, and the Bayes factor it yields may or may not be close to the Bayes factor resulting from the researcher's prior. In most cases, BIC is more conservative than the alternative priors proposed by Weakliem and other critics of BIC, in the sense that the alternative priors are more likely to find evidence for the effect being studied. In practice, therefore, there seems to be some agreement that BIC is sufficiently conservative, perhaps too conservative. It follows that, in most cases, if BIC finds evidence for an effect, we should agree that the data support the effect.

In this article authors discusses over the critical article, by David L. Weakliem, on the Bayesian Information Criterion (BIC) for Model Selection, contained in this issue of the journal "Sociological Methods & Research" itself. Authors report that Weakliem's article makes three main points. First, that the BIC corresponds to a specific probability model that, in many examples, does not make good scientific sense. Second, that the mathematical form of the BIC can lead to pathologies with respect to sample size, data grouping, and model formulation, as illustrated with several real examples in Weakliem's article. Third, that using the BIC to select a model leads to scientifically misleading conclusions. This third point is a problem with any automatic model selection method. One can generally use both analytical and empirical approaches to evaluate applied statistical methods, but empirical evaluations and narratives of applications are suggestively more important determinants of how useful researchers view a statistical method to be in a specific application context.

This article presents a commentary note on the critical article, by David L. Weakliem, on the Bayesian Information Criterion (BIC) for Model Selection, contained in this issue of the journal "Sociological Methods & Research" itself. According to the authors of this article, Weakline has exposed some fundamental problems with the use of the BIC and with "automatic" model selection criteria generally. As is pointed out by Weakliem, a Bayes factor or the BIC, suggestively, cannot tell which statistically significant effects are important in relation to substantive questions of interest. The model or models most appropriate to the purposes of an investigation will typically depend on what those purposes are. In particular, the existence and quantification of small effects, which for some purposes might be best neglected, are often a focus of substantive interest, perhaps especially in studies involving large amounts of data. For judging the substantive importance of effects found to be "real" by a test of significance, responsibility rests with the investigator. Bayesian hypothesis-testing paradigm revolves around a probability, the probability that the effect under study is nonzero; subject-matter importance of the effect plays no part in the motivation or calculation of such probability. Thus, when a Bayes factor or the BIC succeeds in identifying an effect of substantive importance, such success are supposed to be regarded as just a positive luck.

This article presents a critical note on the Bayesian information criterion (BIC) which has recently been proposed as an effective method of model selection in sociological research. The BIC is defined as L2-rlog(N), where "N" is the total sample size and "r" is the number of degrees of freedom remaining after fitting the model, with smaller values representing better fits. Because it includes an adjustment for sample size, the BIC often favors a simpler model even when that model would be rejected by standard hypothesis tests. The BIC was initially proposed for log linear models, and has been employed in most articles using log linear or log multiplicative models published in the leading American sociology journals in the 1990s. The BIC is based on a Bayesian approach to hypothesis testing. In this approach, the observed data and prior beliefs about the probable values of the parameters are combined to produce a Bayes factor; that is, a measure of the weight of evidence that the data provide for or against the null hypothesis. In general, different prior beliefs produce different Bayes factors, and in principle anyone with access to the data could compute the Bayes factor that follows from his or her own beliefs.

The article focuses on statistical techniques and sociological theory. There has always been a close association between the development of quantitative data analysis in sociology and a theoretical strategy that can be described as positivist or neopositivist. But this is a contingent and not a necessary association. Statistical analysis of social data does not necessarily implicate or support the positivist version of uncovering general laws of society. The technical development of statistical analysis in nineteenth- or twentieth-century social science was not animated principally by a theoretical commitment to positivism. The rationale for uncovering social laws with the aid of statistical methods has a theoretical and not a technical basis. It appears in proposals for increasingly formal modes of theory construction, and it presumes the existence of invariant social laws that can be discovered by counterfactual testing of operationalized propositions that are cast in the logic of concomitant variation. Statistical research in sociology is guided as much by this theoretical strategy as by the technical conventions of multivariate analysis.

A linear structural equation model (SEM) without free parameters has two parts: a probability distribution and an associated path diagram corresponding to the causal relations among variables specified by the structural equations and the correlations among the error terms. This article shows how path diagrams can be used to solve a number of important problems in structural equation modeling; for example, How much do sample data underdetermine the correct model specification? Given that there are equivalent models, is it possible to extract the features common to those models? When a modeler draws conclusions about coefficients in an unknown underlying SEM from a multivariate regression, precisely what assumptions an being made about the SEM? The authors explain how the path diagram provides much more than heuristics for special cases; the theory of path diagrams helps to clarify several of the issues just noted. [ABSTRACT FROM AUTHOR]