• Department
  • Undergraduate Program
  • Graduate Program
  • People
  • Research
  • GIVE
  • UC Davis Psychology Quantitative Program

    Quantitative Brown Bag Series

    Location:  Young 166 (unless specified otherwise)
    Thursdays 1:30pm - 2:20pm
    Edit BB Events

    Calendar Administrator:   


    ACADEMIC YEAR:      2013 - 2014 Print Page
    Fall 2013
    SPEAKER: Joseph Gonzales
    Construct Validity of Risk Indices: Analytic Techniques, Costs and Benefits, and Two Examples 
    SPEAKER: Dr. Keith Widaman
    Evaluation of a Bayesian Approach to Estimating Nonlinear Mixed-Effects Mixture Models 
    SPEAKER: Dr. Kevin Grimm
    SPEAKER: Jonas Miller
    Pattern-Mixture Two-Part Mixed Models for Semi-Continuous Data with Nonignorable Missingness 
    SPEAKER: Dr. Shelley Blozis
    SPEAKER: Melissa McTernan
    SAS PROC NLMIXED was developed for the estimation of nonlinear mixed models, including both multilevel and longitudinal data. Nonlinear mixed models are models for which a random coefficient may enter an equation nonlinearly. PROC NLMIXED may be used to estimate models for which a response follows a normal, binomial, or Poisson distribution. For other distributions, the program includes an option for the user to express a likelihood function. This naturally allows for more complex distributions, including joint distributions between variables that follow possibly unique distributions. Thus, this is an advantage to researchers because it allows for a great variety of response functions. In this brief talk, I will describe steps to developing a likelihood function and corresponding NLMIXED syntax to estimate the model. Empirical examples are provided, including specification of a joint longitudinal model for a normal and binomial response.
    SPEAKER: Katerina Marcoulides
    Linking Attentional Training and Performance Across Scales of Time and Methods of Measurement 
    SPEAKER: Stephen Aichele
    Linking Attentional Training and Performance Across Scales of Time and Methods of Measurement The dynamics of psychological processes, as they unfold within individuals over time, are fundamental to understanding behavioral, emotional, and cognitive development. The incorporation of multiple methods of assessment, and their application across different time scales and settings, is essential to the advancement of empirical knowledge through psychological research. But as scientists examine increasingly complex relations between psychological variables, there is an ever greater need to identify potential sources of bias during analysis. Data aggregation within variables (e.g., across persons, attributes, and/or settings) influences patterns of association between variables, and this in turn affects research outcomes. Furthermore, different analytic objectives (e.g., causal inference, valid interpretation of within-person change) may conflict due to experimental and/or statistical constraints. In the present work, I illustrate different strategies for data aggregation and analysis in two observational studies from the Shamatha Project (SP), a longitudinal investigation of intensive meditation practice. The general aim of these studies was to examine person- and training-specific variables in relation to previously reported experimental outcomes. Implications for quantitative theory are discussed in light of these studies' results.
    The Relation of Depressive Symptoms and Socioeconomic Status on Diabetes Self-Care Management and BMI in Rural Latinos with Diabetes 
    SPEAKER: Marilu Isiordia
    This study explored the relationships between depressive symptoms and socioeconomic status (SES) on self-care management (e.g., diet and physical activity) and body weight in 250 rural Latinos with diabetes. I hypothesized there would be a negative correlation between self-reported depressive symptoms and ratings of adherence to (a) healthy diet, (b) physical activity, and (c) a positive correlation with body mass index (BMI). I also hypothesized that SES would have (d) a positive correlation with healthy diet, (e) a positive correlation with physical activity, and (f) a negative correlation with BMI. Additionally, I conducted exploratory analyses to examine the influence of neighborhood perception on diabetes self-care management and BMI. Significant results indicated depressive symptoms were negatively correlated with diabetes self-care management. Contrary to these predictions, negative relationships between annual income and both healthy diet and physical activity were found. Results also showed a positive relationship between education and physical activity and BMI. Finally, neighborhood perception was positively correlated with self-care behaviors and negatively correlated with BMI. Overall, these findings help indicate how the presence of depressive symptoms and SES are associated with diabetes self-care management in Latinos with diabetes.
    SPEAKER: No brown bag
    Winter 2014
    SPEAKER: Michelle Harris
    Global self-esteem is thought to stem, at least partly, from positive parent-child relationships. Previous cross-sectional research supports this hypothesis, but longitudinal studies have failed to find robust evidence of prospective effects. The current study uses data from Germany and the United States to test the prospective relation between parent-child closeness and adolescent self-esteem development. We tested this relation using self, parent, and observer reports of parent-child closeness and self-reported self-esteem from ages 12-16. Results replicated the concurrent correlation, but six types of longitudinal models (regressions, autoregressive cross-lagged models, latent growth curve models, latent profile analyses, latent difference score models, and enduring effects vs. revisionist models) failed to show a prospective relation. Thus, the longitudinal effect of parent closeness on self-esteem is, at best, small and difficult to detect. These findings call into question the purported impact of parent-child relationships on self-esteem, at least during the adolescent years.
    Timing is everything: Ovulation estimation method efficacy 
    SPEAKER: Joseph Gonzales
    Ovulatory effects research in humans is typically conducted by comparing time points within persons that are associated with critical periods of the ovulatory cycle. Specifically, researchers compare observational and self-report data corresponding with the late-follicular and mid-luteal phases to test for shifts corresponding to reproductive fertility or hormonal processes. Critical to this research design is the estimation of ovulation in order to sample correctly from target phases. Typical approaches include the forward and backward counting methods, and hormone detection methods (e.g., LH urine testing). While the forward counting method has largely been dismissed as unreliable due to evidence of follicular phase variability between- and within-women, variability in the still popular backward counting method may also require larger sampling in order to overcome phase-sampling error. In the present study I present four representations of behavioral shifts associated with the ovulatory cycle and evaluate the efficacy of conventional sampling methods with respect to detecting differences between late-follicular and mid-luteal phases.
    Estimation of Factor Scores from Multiple-group Item Response Models: Implications for Integrative Data Analysis 
    SPEAKER: Pega Davoudzadeh
    Data integration refers to obtaining multiple data sets, scaling their measurements, and analyzing them as though they represent a single data set. A first step is to scale measurements to common scales, which is typically done with multiple group item response models. Latent variable scores are then estimated and used in subsequent analyses. This approach was found to produce inconsistencies in latent variable estimates for individuals from different studies with the same response pattern. Monte Carlo Simulations were then conducted to evaluate the accuracy of latent variable estimates from this and other approaches that ignore the nesting of participants in studies. Results suggest that ignoring study differences led to slightly more accurate latent variable estimates that were also consistent. This approach is recommended for data integration and scaling of latent variable scores across studies. Implications for longitudinal data integration are discussed.
    My answer to Tukey's question: Does anyone know when the correlation coefficient is useful? 
    SPEAKER: Dr. Fushing Hsieh
    I will begin with two real examples to illustrate association patterns in bipartite network data. Then the final example is about the drug-gene-cell*line triangular association relationship. The asymmetry of such a relationship reveals where correlation coefficient does and doesn't work
    SPEAKER: No Brown Bag
    Nonlinear Mixed Models: A SAShay 
    SPEAKER: Nathan Smith
    I will be using the PROC NLMIXED option in SAS to compare different estimation techniques (Gaussian Quadrature, First Order Methods, Importance Sampling, and Hardy Quadrature). With data from an empirical study and utilizing a structured latent curve model to fit the data, I test the different options available in SAS for PROC NLMIXED. After providing background for both the empirical data and methods used in my analysis, I will present my findings and introduce a number of questions arising from the differences in results.
    SPEAKER: Dr. Jon Helm
    Quantitative Methods for Agent-Based Models 
    SPEAKER: Matt Miller
    I will discuss some of the goals and results of agent-based models (ABMs) used in psychology and other disciplines in relation to the challenges of characterizing ABM results quantitatively. I will offer some thoughts and solicit suggestions on appropriate quantitative methods for ABMs.
    From Functional to Neuroimaging Data 
    SPEAKER: Dr. Jane-Ling Wang
    Functional data are random functions on an interval or set and have emerged frequently in modern scientific experiments. The analysis of such data is termed "Functional Data Analysis (FDA)" in the statistical literature. In this talk, we first give a brief overview of the various types of functional data and the various approaches to handle them. Then we show how the FDA approaches could be useful to analyze neuroimaging data, which are intrinsically functional data. Two applications will be illustrated, one for PET time course data and the other for fMRI time series.
    Statistics for Individual Change Assessment and their Relationship with Group Change 
    SPEAKER: Eduardo Estrada
    I will describe and briefly discuss some simple statistics for assessing individual change and their relationship with classical group change statistics such as T test and effect size measures.
    Nonignorable Nonresponse, Random Subsamples of Non-respondents and How Best to Adjust for the Missing Data 
    SPEAKER: Dr. Christiana Drake
    Early School Readiness Predictors of Grade Retention from Kindergarten through Eighth Grade 
    SPEAKER: Pega Davoudzadeh & Melissa McTernan
    SPEAKER: Dr. Nilam Ram
    SPEAKER: Melissa McTernan
    SPEAKER: Jimmy Calanchini
    SPEAKER: Chelsea Muth
    SPEAKER: No brown bag
    SPEAKER: Carolyn McCormick