In this, our first volume of
Sociological Methodology, we have grouped the articles into four sections into which the papers naturally seemed to fit. The first section is entitled Statistical Models: Prediction, Model Evaluation, and Missing Data. Papers in this section build on the long-standing tradition in sociology that relies on statistical models based on observational data, and they focus on important issues in prediction, model evaluation, and missing data. In On the Assignment of Individuals to Latent Classes, Leo A. Goodman compares two main competing procedures as to how individuals in a multiway contingency table can be assigned to latent classes in a latent class analysis. Andrew Gelman and Iain Pardoe, in Average Predictive Comparisons for Models with Nonlinearity, Interactions, and Variance Components, discuss the problem of predictions from regression models that are nonlinear, multilevel, and interactive. In Multilevel Covariance Structure Analysis by Fitting Multiple Single-Level Models, Ke-Hai Yuan and Peter M. Bentler propose methods for estimating and evaluating structural equation models at separate levels for multilevel data. Finally, in Regression with Missing Ys:AnImproved Strategy for Analyzing Multiply Imputed Data, Paul T. von Hippel offers concrete and sensible advice when dealing with missing data on the dependent variable in regression analysis.
The second section of the volume focuses on data issues, with the title of Data Collection and Data Quality. In this section, we include two papers. In Calibrating Measures ofFamily Activities between Large-and Small-Scale Datasets, Nora Broege, Ann Owens, Anthony P. Graesch, Jeanne E. Arnold, and Barbara Schneider compare data from one large survey and data from one in-depth study on time use and activities in families. In Extensions of Respondent-Driven Sampling: Analyzing Continuous Variables and Controlling for Differential EDITOR S INTRODUCTION xix Recruitment, Douglas D. Heckathorn extends his earlier work (with collaborators) on respondent-driven sampling to the case of continuous variables and controlling for differential recruitment.
Social network is one of the key research areas in contemporary sociology.We are proud to present three methodological contributions to network analysis in this volume, under the section heading of Methods for Analyzing Social Network Data. In Very Local Structure in Social Networks, Katherine Faust asks the important question of to what extent empirical social works can be accounted for by simple local structural properties, and her answer is a lot. The other two papers are authored by Carter T. Butts, enjoying the rare luxury of publishing two papers in a single issue of a journal, which in our case is also a single volume! In Permutation Models for Relational Data, Butts proposes a class of models for measuring the association among dyadic structures. In Models for Generalized Location Systems, he proposes a class of broad, theoretically innovative models that are potentially usable for studies in many areas, including occupational stratification and residential segregation.
Finally, we continue with the theme of the 2005 volume, which featured an important contribution by James Heckman, on causal inference in social science. The present section on the same topic contains two papers. Indices ofRobustness for SampleRepresentation by Kenneth Frank and Kyung-Seok Min expands the notation of robustness from the traditional concern with omitted variables to concerns with out-of-scope extrapolation. In Identification and Estimation of Causal Effects with Time-Varying Treatments and Time-Varying Outcomes, Jennie E. Brand and Yu Xie extend the counterfactual framework in causal inference to a longitudinal setting, focusing on the problem of the uncertainty reference groups, and they propose a class of estimands with a forward-looking principle.
