|
| January 25 |
|
Recruitment Day |
| February 1 |
|
TBA
Vince Staggs |
| February 8 |
|
Pre-Post Designs
Emily Fall |
| February 15 |
|
Growth rate modeling techniques for longitudinal data
Johnny Zhang
In this talk, I will present models I am developing to analyze rates of growth. After discussing the significance of rates of growth in understanding dynamic processes, I will present the simple growth rate models built on the first derivative of growth curves and the compound growth rate models built on the latent difference score models. The use of the models will be demonstrated by analyzing children's mathematical performance data. Finally, simulation studies will be conducted to evaluate the performance of the models.
|
| February 22 |
|
Mediation analysis: Three problems with current practice
Kristopher J. Preacher
The statistical analysis of mediation effects has seen many remarkable advances over the last 25 years, but many challenges remain. This presentation will describe-and suggest solutions to-three problems that arise from following expert recommendations. First, researchers are often advised to use structural equation modeling (SEM) to explore mediation effects. However, different approaches to identifying the model can lead to hypothesis tests that yield different results. The most common method of model identification leads to the least powerful test of mediation. Second, bootstrapping has emerged as a powerful way to test mediation hypotheses, but the most commonly recommended bootstrapping methods suffer from a fundamental flaw-the null hypothesis of no mediation is not tested. Third, the way in which mediation effects are operationalized is not consistent with how mediation is classically defined. These problems are discussed, and solutions are described.
|
| February 29 |
|
The new Quant.KU website
Don Gay
|
| March 7 |
|
No Meeting
|
| March 14 |
|
Introduction to power analysis: Concept, terms, and what you will need to conduct your first basic power analysis
Jason Cole
Statistical power is a term that describes the likelihood of finding a significant result in your sample if the result is significant in the entire population. Therefore, we use power analyses to either determine how much power we have given our sample size (less common) or how many subjects we need to achieve a desired power level (more common). The current lecture will review the concept of power analysis and sample size estimation, discussion of terms used in power analysis, and a walkthrough of everything you'll need to know to conduct your first power analysis. As an added benefit, to conduct a proper power analysis you will need to plan your study with careful and detailed understanding.
|
| March 21 |
|
Spring Break (no meeting)
|
| March 28 |
|
Training for Use of Computing Cluster
Phil Hauptman |
| April 4 |
|
Discussion of Ethics
Waylon Howard |
| April 11 |
|
SAS IML Program for Combining Estimates from Multiple Imputations
John Geldhof |
| April 18 |
|
Summit on Teaching Undergraduate Statistics
Gita Sawalani & James P. Selig |
| April 25 |
|
TBA
Jason Lee |
| May 2 |
|
Growth Curve Modeling of Language Indicators
Libby McConnell |
| May 9 |
|
Stop Day (no meeting)
|
| August 24 |
|
Planning |
| August 31 |
|
Missing data [PPT file]
Todd Little
This presentation will cover the types of missing data mechanisms (MCAR, MAR, & NMAR), how they may arise, and, most importantly, what to do when you have missing data. I will describe a convincing rationale for why missing data imputation is NOT cheating and why you should impute missing data in nearly every circumstance! This talk assumes little or no statistical background. |
| September 7 |
|
Discussion of SAS data functions [PPT file] [SAS file]
Ihno Lee |
| September 14 |
|
Discussion of basic SAS procedures [SAS file]
Mike Clark |
| September 21 |
|
Discussion of SAS PROC MIXED and effect size estimation in multilevel modeling [PPT file]
Vince Staggs |
| September 28 |
|
Discussion of missing data and SAS PROC MI [PPT file] [SAS files]
James Selig and John Geldhof |
| October 5 |
|
Measurement invariance
Todd Little
This presentation will cover the concept of factorial invariance (also known as measurement equivalence). Factorial invariance is assumed any time one compares across two or more groups or across two or more time points. This assumption MUST be tested in order to draw valid conclusions in such situations. I will cover the theoretical, conceptual, and mathematical basis for the conditions under which factorial invariance should hold and conditions under which it will not. I will also describe the sequence of steps that are needed to test the assumption of factorial invariance. Although I will describe factorial invariance in the context of structural equation modeling, I do not assume that one has extensive knowledge of SEM, just a basic/conceptual understanding. |
| October 12 |
|
Fall break (no meeting) |
| October 19 |
|
SMEP (no meeting) |
| October 26 |
|
Teaching evaluations
Gita Sawalani |
| November 2 |
|
Model identification and partial invariance [PPT file]
Kris Preacher
I present a Monte Carlo simulation study that compares the results of tests of partial measurement invariance across three methods of scale-setting. Two methods of scale setting have dominated the literature on invariance testing: the marker variable approach, in which one loading per factor is fixed to 1.0, and the reference-group approach, in which the factor mean and variance are fixed to 0.0 and 1.0 (respectively) in one group and estimated freely in the remaining groups. Recently, Little, Slegers, and Card (2006) introduced an effects-coding constraint approach, in which the mean item intercept for each factor is fixed to 0.0 and the mean loading is fixed to 1.0. We examine the consequences of imposing the different scaling and identification conditions represented by these approaches for judgments of partial measurement invariance. Conditions likely to impact conclusions are varied, including sample size, effect size, and invariance testing strategy. |
| November 9 |
|
Differential item functioning [PPT file]
Jason Lee |
| November 16 |
|
Bootstrapping [PPT file]
James Selig |
| November 23 |
|
Thanksgiving break (no meeting) |
| November 30 |
|
Validation procedures for cluster analysis
Emily Ledford |
| December 7 |
|
Stop Day (no meeting) |
| January 26 |
|
Planning |
| February 2 |
|
Discussion of John R. Nesselroade's research [selecting indicators]
Todd Little and James Selig |
| February 9 |
|
Discussion of John R. Nesselroade's research [adult personality development]
Vince Staggs |
| February 16 |
|
Discussion of John R. Nesselroade's research [idiographic filters]
Emily Ledford |
| February 23 |
|
Discussion of John R. Nesselroade's research [SMEP presidential address]
Mike Clark |
| March 2 |
|
APA midwinter meeting (no meeting) |
| March 9 |
|
Discussion of John R. Nesselroade's research [Chow & Nesselroade,
2004]
John Geldof |
| March 16 |
|
Discussion of John R. Nesselroade's research [Ram & Nesselroade,
in press]
Libby McConnell |
| March 23 |
|
Spring Break (no meeting) |
| March 30 |
|
SRCD (no meeting) |
| April 6 |
|
Median Splits and Extreme Groups
Kris Preacher
I examine the practice of dichotomization of quantitative measures, in which relationships among variables are examined after one or more variables have been converted to dichotomous variables by splitting the sample at some point on the scale(s) of measurement. A common form of dichotomization is the median split, where the independent variable is split at the median to form high and low groups, which are then compared with respect to their means on the dependent variable. A related technique is the extreme groups approach (EGA), in which analysis of continuous variables is preceded by selecting individuals on the basis of extreme scores of a sample distribution and submitting only those extreme scores to further analysis. EGA is often used to achieve greater statistical power in subsequent hypothesis tests. However, there are several largely unrecognized costs associated with both dichotomization and EGA that must be considered, including effects on power, standardized effect size, reliability, model specification, and the interpretability of results. These practices are rarely defensible and often will yield misleading results. |
| April 13 |
|
Distribution of Cronbach's Alpha for Ordinal Data: A Bayesian Based Approach
Byron Gajewski (KU Medical Center)
The author proposes to obtain point estimates and intervals for Cronbach's alpha when the data are ordinal. Traditional calculations of Cronbach's alpha on ordinal instruments underestimate the true Cronbach's alpha provided from the latent variables that are assumed to produce the ordinal data. By utilizing Bayesian models and data augmentation, confidence (credible) intervals are provided for Cronbach's alpha at the latent variable level. The proposed methodology is shown to have theoretically correct coverage probability and is demonstrated on an instrument that measures nursing home residents' quality of life. |
| April 20 |
|
Discussion of John R. Nesselroade's research [Warp and Woof of Developmental Fabric]
James Selig |
| April 27 |
|
Discussion of John R. Nesselroade's research [Sampling and Generalizability]
Waylon Howard and Jason Lee |
| May 4 |
|
Early Stop Day (no meeting) |
| May 11 |
|
Discussion of John R. Nesselroade's research
John R. Nesselroade |
| August 18 |
|
Planning |
| August 25 |
|
Program discussion |
| September 1 |
|
Maximum Likelihood Estimation
Vince Staggs |
| September 8 |
|
Using the Direct Autoregressive Factor Score (DAFS) Model to Examine the Relations between Sleep and Mood in Children
James Selig |
| September 15 |
|
Profiles of Literacy Skills
Emily Ledford
Over 315 adult education participants completed a comprehensive battery of reading skills assessments. The battery provided assessment of five recognized reading components. Cluster analysis using participants' scores on these reading components yielded profiles of unique learner needs and recommendations for instructional activities to address those literacy needs. |
| September 22 |
|
(no meeting) |
| September 29 |
|
Practice Makes Perfect: Using a Proportional Odds Model to Determine if Number of Homework Assignments Completed Determines Success on the Final Exam in a Psychology Statistics Course
Gita Sawalani
The current study investigates whether or not number of homework assignments completed affects a student's performance on the final exam for an undergraduate psychology statistics course. Not a lot of research on homework affecting academic performance has been done (Swank, 1999). The null hypothesis states that the number of homework assignments completed does not have any effect on one's grade on the final exam for the course. The subjects for this study included 90 students who took the psychology statistics course at the University of Kansas in the fall semester of 2004. A multilogit regression framework was used to fit a proportional odds model. Results suggest that there is strong evidence that number of homework assignments completed predicts one's grade on the final exam for the course. |
| October 6 |
|
Testing, Plotting, and Probing Interactions
Kris Preacher
Main effects in additive models are often insufficient to characterize the effect of a predictor variable on a dependent variable. In many circumstances it can be shown that the effect of X on Y differs across groups, or varies as a function of at least one moderator variable. This teaching talk will explore statistical and visual strategies for exploring and illustrating 2-way and 3-way interaction effects in the ANOVA and regression frameworks. |
| October 13 |
|
Fall Break (no meeting) |
| October 20 |
|
SMEP (no meeting) |
| October 27 |
|
New MIMIC model for DIF identification
Jason Lee |
| November 3 |
|
Using Cognitive Diagnosis Models to Analyze and Diagnose Psychological Disorders
William P. Skorupski (KU Department of Psychology and Research in Education)
The field of psychometrics is extremely important to any social science researcher or practitioner who desires to obtain and use measurements of some human construct. Item response theory (IRT) is set of psychometric techniques that provides many applications for scaling assessments and conducting research in the social sciences. The purpose of this talk will be to introduce the main concepts and models used in IRT, consider ways in which IRT can be applied to practical assessment problems, and acquaint the audience with some interesting research areas in IRT. |
| November 10 |
|
The Role of the Quantitative Psychologist in Clinical Trials [PPT file]
Janet Levy (Center for the Clinical Trials Network, National Institute on Drug Abuse, Bethesda, Maryland)
The purpose of the seminar is to educate faculty and students about exciting career opportunities in clinical trials. Following a brief introduction, the role of the quantitative psychologist in both the design and implementation of clinical trials will be discussed. An emphasis will be placed on the links between the scientific rationale, the clinical questions posed, and statistical hypotheses in the context of clinical research. It will be argued that quantitative psychologists are particularly well trained to make these intellectual links. Two examples from the author's recent experience will be described to illustrate the contributions that a quantitative psychologist can make in the design of clinical trials in addiction. Some emerging methodological issues in clinical trials will also be discussed. Finally, the author's personal educational and professional experiences will be described, emphasizing the contributions of both psychology and mathematics coursework to current work in clinical trials. |
| November 17 |
|
Basics of SAS Programming [SAS files]
Todd Little |
| November 24 |
|
Thanksgiving (no meeting) |
| December 1 |
|
New Features and Unresolved Issues in Traditional Longitudinal (Panel) SEM
[PPT file] [Excel file]
Todd Little
This talk will focus on the traditional longitudinal panel design and the use of SEM to analyze them. Some new features to be discussed include new methods for scaling and identification of constructs, and testing and evaluating factorial invariance across time. Unresolved issues include the inclusion and treatment of covariates and inferring causality. |
|
|
