WPA Forms - Statistics Workshops Registration Form

WPA 2010 Statistics Workshops

A statistics workshop will be held each day of the WPA convention in Cancun. Details of each workshop are described below. The workshops require a separate registration. Fill in the information at the bottom of this form to create a mail-in registration form.

Thursday 2:15 – 4:15
Testing Moderation Hypotheses using Structural Equation Modeling (SEM) Techniques
Presenter: Jodie Ullman, CSU San Bernardino
Chair: TBA

SEM techniques offer rich methods to test all different types of moderation (interaction) hypotheses.  Through the use of multiple group structural equation models it is possible to test whether or not a variable say, ethnicity or gender, actually moderates, not just regression paths or means, but also the composition of the factors/constructs themselves. For example, does the composition of a construct like acculturation change as a function of ethnicity or does a construct of risky sex behavior change as a function of gender?  

In this workshop we begin by exploring interactions using measured and latent variables in standard single group structural equation models. After a fairly brief exploration of these models we will turn to multiple group SEM models and look at a variety of types of moderation/interaction hypotheses. Examples using a variety of structural equation modeling software packages will be provided including EQS, AMOS, and MPlus.

Friday 8:30 – 10:30
Structural Equation Modeling Approach to Classical Reliability Theory and Guttman Scaling
Presenter: Peter M. Bentler, UCLA
Chair; TBA

This workshop will overview reliability coefficients based on structural equation modeling, including coefficient alpha, a coefficient based on the assumption of unidimensionality of variables/items, a coefficient based on any structural model for variables, and two greatest lower bound coefficients making minimal assumptions about the structural model. Optimally weighted coefficients are discussed if time permits. Based on a modern update to Bentler (1971), a new structural equation approach to building and evaluating scales based on binary Guttman data is presented. This approach provides an alternative to the Rasch model, but as in traditional structural modeling, it allows model testing via chi- square as well as generalization to more complex model structures when more than one parameter per item is needed.

Saturday 8:45- 10:45
A Workshop on Quasi-Experimental Design
Presenter: William D. Crano
Claremont Graduate University
Chair: TBA

Far from the court of last resort, quasi-experimental designs provide the creative researcher with a host of options that often are more desirable than that the randomized experiment makes available. The quasi-experiment can go where the experiment fears to tread, and can provide useful information when implementing experimental designs is impossible. The drawback of all quasi-experimental designs, the incapacity to assign participants randomly to study condition, is an admitted inconvenience, but it is not a death knell. Many designs are available to the quasi-experimentalist that provide evidence whose causal implications are so compelling that they are practically undeniable. Indeed, the causal inferences that may be drawn from some quasi-experimental designs are at least as certain as those available from the classic experimental design – their advantage is that the quasi-experiment could be performed, whereas an experiment could not given the constraints of the research context. In this workshop we will explore a variety of subtypes of quasi-experimental designs, ranging from simple and complex interrupted time series studies to nonrandomized pretest/posttest/control group designs to regression discontinuity approaches. Real research problems that do not lend themselves to the simple expedient of the randomized assignment of subjects to conditions will be discussed, and the audience will be invited to solve them while learning about the possibilities made available in the class of designs described as quasi-experimental.

Sunday 8:30 – 10:30
Introduction to Binary Logistic Regression and Propensity Score Analysis
Presenter: Dale E. Berger, Claremont Graduate University
Chair: TBA

Psychologists often wish to use multiple predictors to predict or model a dichotomous outcome (e.g., success/failure, persist/dropout, admit/deny, self selection into a treatment vs. control as in propensity analysis). Ordinary regression does not provide an appropriate model for this type of analysis, but logistic regression is a readily available alternative that is accessible in SPSS/PASW and other statistical packages. Logistic regression is not difficult to use and understand although new terminology and unfamiliar statistics can be challenging for first-time users. In this demonstration we will examine the logic and application of logistic regression for dichotomous dependent variables, show why ordinary regression is not appropriate, and demonstrate applications with dichotomous predictors, continuous predictors, and categorical predictors. When people self-select into a treatment program, analysts may attempt to form comparison groups through blocking or through analysis of covariance. Another option is to use propensity score analysis whereby logistic regression is used to compute the propensity for someone to self-select into either a treatment group or a control group. People then can be blocked into comparison groups with similar propensity for selecting the treatment condition, perhaps providing satisfactory control of variables associated with self-selection. Participants in the workshop will be given a packet with SPSS/PASW syntax and annotated output for a range of applications, including an introduction to propensity score analysis. Familiarity with multiple regression analysis will be helpful, but not required.

Registration

Fill in the information below. When you submit the information, a form will be created that you can print and mail in with your check. Registration for each workshop is $5 for students, $10 for professional members. You must also register for the convention to attend the workshops.

Registration Type:	Professional Student
Name:
Affiliation:
E-mail Address:
Address:
City:
State:
Zip:
Daytime Phone:
Check Workshops Desired
	Thursday 2:15 – 4:15 Testing Moderation Hypotheses using Structural Equation Modeling (SEM) Techniques
	Friday 8:30 – 10:30 Structural Equation Modeling Approach to Classical Reliability Theory and Guttman Scaling
	Saturday 8:45- 10:45 A Workshop on Quasi-Experimental Design
	Sunday 8:30 – 10:30 Introduction to Binary Logistic Regression and Propensity Score Analysis