When a linear structural equation model results in nonlinear procedures for solution, multiple (but finite) sets of solutions may be possible, even with population data. An example of a three-variable feedback system is given in which two sets of solutions may be generated. (Author/CTM)

Two techniques for interpretation of fitted log-linear models in contingency table analysis are discussed. The use of odds ratios as opposed to direct interpretation of the fitted model is argued for. (Author/JKS)

A recent recommendation by Holt (EJ 200 576) that coefficients resulting from estimating log-linear and similar models should not be interpreted is argued to be based on lack of clarity about the substantive and theoretical importance of the choice between dummy and effect coding for categorical variables. (Author/GDC)

Two methods of analyzing nonorthogonal (uneven cell sizes) cross-classified data sets are compared. The methods are direct standardization and the general linear model. The authors illustrate when direct standardization may be a desirable method of analysis. (JKS)

Four techniques for assessing differences between least-squares estimators of regression coefficients from group and individual-level data are summarized. The utility and suitability of all four approaches are discussed when (1) data are grouped by a nominal characteristic; (2) there are multiple regressors; or (3) individual-level data cannot be…

This paper poses a research question about how a person's work experience affects his or her earnings and shows how the Cumulative Experience Method (CEM) can provide an answer to the question using all available information in a longitudinal data set. (Author/GDC)