Joseph E. Gonzales

Efficacy of methods for ovulation estimation and their effect on the statistical detection of ovulation-linked behavioral fluctuations

Contention of the ovulatory shift hypothesis is principally supported by failures to replicate previous findings; e.g., recent meta-analytic work suggests that the effects endorsing the hypothesis may not be robust. Some possible limitations in this and other ovulatory-effects research—that may contribute to such controversy arising—are: (a) use of error-prone methods for assessing target periods of fertility that are thought to be associated with behavioral shifts, and (b) use of between-subjects—as opposed to within-subjects—methods. In the current study we present both simulated and empirical research: (a) comparing the ability of between- and within-subject t-tests to detect cyclical shifts; (b) evaluating the efficacy of correlating estimated fertility overlays with potential behavioral shifts; and (c) testing the accuracy of counting methods for identifying windows of cycle fertility. While this study cannot assess whether the ovulatory shift hypothesis or other ovulatory-based hypotheses are tenable, it demonstrates how low power resulting from typical methods employed in the extant literature may be associated with perceived inconsistencies in findings. We conclude that to fully address this issue greater use of within-subjects methodology is needed.

Individual Pooling for Group-Based Modeling Under the Assumption of Ergodicity

Psychology principally utilizes nomothetic, interindividual approaches to model phenomena of interest. However, it is the case that these approaches do not always capture the processes for each individual in the sample. If the research is focused on individual processes, confining analysis to the idiographic level may be more appropriate. One way to overcome the nomothetic inability to capture idiographic processes is to identify those participants who meet the criteria of ergodicity and restrict analysis to the resulting sample. Under these conditions it is quantitatively justifiable to create a group model without concern that it may fail to represent each members idiographic process. In this study we explore the utility of such a method by (a) applying an ergodic pooling test to a sample of dyads (N = 128) who provided daily (T = 50) self-reports of affect, (b) applying an ergodic pooling test to samples (N = 4) of simulated ergodic time series data (T = 50, 250, and 1,000), (c) modeling dyads and simulated subgroups identified as ergodic, and (d) comparing the results from a model specified at the group level with those from models specified at the individual level.

Harsh parenting, physical health, and the protective role of positive parent-adolescent relationships.

RATIONALE Harsh, abusive and rejecting behavior by parents toward their adolescents is associated with increased risk of many developmental problems for youth. OBJECTIVE In the present study we address behaviors of co-parents that might help disrupt the hypothesized health risk of harsh parenting. METHOD Data come from a community study of 451 early adolescents followed into adulthood. During early adolescence, observers rated both parents separately on harshness towards the adolescent. Adolescents reported on their physical health at multiple assessments from age 12 through age 20, and on parental warmth. RESULTS Harsh parenting predicted declines in adolescent self-reported physical health and increases in adolescent body mass index (BMI). Although the health risk associated with harshness from one parent was buffered by warmth from the other parent, warmth from the second parent augmented the association between harshness from the first parent and change over time in adolescent BMI. CONCLUSION As appropriate, preventive interventions should include a focus on spousal or partner behaviors in their educational or treatment programs. Additional research is needed on the association between self-reported physical health and BMI in adolescence.

Longitudinal Models for Studying Multivariate Changes and Dynamics

In this paper, we describe a longitudinal modeling approach for examining multivariate changes and dynamics. This technique is based on latent change scores and is executed using a structural equation modeling framework. We provide an overview of the model, describing desirable features for identifying dynamics among multiple processes. We then illustrate its application using empirical data consisting of longitudinal processes and conclude the paper with some potential steps for advancing the modeling possibilities.

Controlling Measurement Heterogeneity in Longitudinal Data Using Covariate Residualized Indicators

Implicit in the modeling of a construct’s latent trajectory is the assumption that strong factorial invariance (i.e., invariant item intercepts and factor loadings) holds across repeated measurements. Yet, many researchers who examine latent means do not test the tenability of this assumption, and when they do, it is unlikely to hold. In this study, we examined whether controlling for the influence of a covariate variable (biological sex) not hypothesized to underlie the intended factor structure, but associated with measurement heterogeneity, could establish strong invariance and recover latent mean changes. For this, we simulated two-occasion longitudinal data consisting of a three-item factor with 250 males and 250 females. At time-two, all item intercepts for males showed measurement heterogeneity of 3, 5, and 7, and amean latent difference of 5. To these data we applied the following covariate residualization techniques: (a) a residualized indicator intercept absent model (RIIAM) in which indicators are formed from the residuals of a linear regression intercept-absent model with the covariate as the predictor; (b) a complete residualized indicatormodel (CRIM) in which indicators are residualized by regressing each itemon the covariate; and (c) a residualized indicator adjustment model (RIAM) in which the linear regression coefficient of the covariate for each indicator is used to weight the covariate scores and are then subtracted from each indicator. The likelihood ratio chi-square difference test and the change in comparative fit index were used to confirm strong invariance was not met in the uncorrected, simulated data, and met after controlling for the covariate source of measurement heterogeneity. To evaluate the accuracy of the residualized approaches compared to uncorrected data, we calculated the relative bias of the latent factor mean at time-two. Results are presented in Table 1. As expected, longitudinal strong invariance did not hold in the uncorrected data across most simulation conditions. However, all three residualized methods

Abstract: Sensitivity to Restricted Factor Analysis Model Misspecification

While there is contention over the use of fit indices (Barrett, 2007; Bentler, 2007), they are often used to assess whether models are tenable representations of data, often utilizing rules of thum...