INSERM U.292 - UNESCOInstitut Federatif de Recherche 69 - Paris, Ile de
FranceRasch Measurement Lecture SeriesDecember 11 - 15 20002:00 to 5:00
pm
David Andrich, PhDSchool of EducationMurdoch UniversityMurdoch, Western
Australia 6150email: [log in to unmask]
Location : Salle de l'IFR, Bat. 15-16, Hopital Paul Brousse,
16 avenue Paul Vaillant Couturier, Villejuif, (metro Villejuif)
Registration fee : Full cost : 5000 FF, Reduced rate :1500 FF
For a registration form, email : [log in to unmask]
INTRODUCTION TO RASCH MEASUREMENT MODELS
(with applications of interactive software)
This is an introductory course for people who (i) have had some
professional experience in latent trait or item response theory and would
like to study the fundamental principles of the Rasch models, and (ii)
those who have had no experience with this item response theory but would
like an introduction to it. The case for applying Rasch models is that
they provide estimates of the item parameters of an instrument which are
independent of the person parameters, and vice-versa. This makes tests of
fit, linking, equating, and testing for differential item functioning,
direct consequences of the model. The course will cover these topics and
applications of the model to both dichotomous and graded response items
found in performance and attitude assessment. The study of the model and
its applications will include working with modern interactive software for
data analysis. Participants are encouraged to bring a data set with them
for analysis. Generally, there should be 10 or more items with responses
that are either dichotomous (e.g. 0 or 1; or 1, 2; or in a multiple-choice
format, A, B, C, D), or graded responses (e.g. 0,1,2,…..m; or 1,2,3,…..m,
where m is the maximum score), and 100 or more people. If the data are in
EXCEL format, it is straightforward to make them ready for analysis. The
specific topics covered in the lectures are described below.
PROGRAM
Lecture 1 : Background: two approaches to measurement.
· The model chosen to fit the data
· The model chosen because it characterizes the criteria for
measurement
Lecture 2 : Dichotomous items: Basic design, structure and reasoning.
· The Guttman structure - deterministic
· The Guttman structure - probabilistic
· The Rasch Model
Lecture 3 : Elementary theory and equations of estimation for dichotomous
items.
· Estimation in dichotomous responses when the parameters are
identical over replications
· Estimation when dichotomous responses are not identical over
replications
· Odds and probabilities
· The structure of the probability matrix and the data matrix
· Examples of linking items where not all persons respond to the same
items
· Conditional probability of a pattern given the total score - the
Guttman pattern again.
· Invariance of parameters in the Rasch model
Lecture 4 : Background statistics and response process for the rating and
partial credit models
· The mean as a theoretical proportion
· The expected value equation
· The kinds of response formats relevant for the model
Lecture 5 : Elementary theory and equations of estimation for items in
rating and partial credit formats
· The structure of the response process
· Parameter estimates
· Test of fit between the data and the model
· Structure of the thresholds defining the ordered categories
· Differential item functioning
Lecture 6 (Optional) : Full derivation of the Rasch model for graded
responses
· The response pattern of latent dichotomous responses
· The mapping of the latent dichotomous responses onto the ordered
categories
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|