Practical two-day workshop
Applied Item Response Theory (IRT) in R
Date: 1 May 2019 - 2 May 2019
Location: David Weatherall Building, Keele University, ST5 5BG
Cost: £295 Early Bird Student, £395 Early Bird Public Sector, £495 Early Bird Commercial/Industry - Early Bird available until 28 February 2019.
Tutors: Led by Dr Gareth McCray, Dr Sara Muller, Kieran Bromley and Professor Gillian Lancaster
Many measurement instruments are questionnaire based and comprise of a number of items with binary (eg. yes/no, pass/fail) or polytomous (eg. always, sometimes, never) responses designed to measure an underlying latent construct eg. child development, pain intensity or depression. Item response theory is an integrated psychometric framework for developing and scoring measures or tests.
This course is for researchers, epidemiologists, psychologists, and statisticians who need to create scores from questionnaire-type measures or who deal with these types of measures. As it is an applied course, it aims to cater for individuals with familiarity in either statistics or testing and measurement.
The course is aimed at individuals that want to learn how to analyse categorical data taken from questionnaire-type measures intended to yield a single score. Familiarity with factor analysis or principal component analysis is not required and these methods will not be covered in the course.
We specifically focus on binary or polytomous clinical response data. We recommend participants have a basic familiarity of applied statistical methods, as well as an understanding of psychometric properties such as reliability and validity.
We also recommend that participants are familiar with R, although the practicals will not require individuals to write their own code. Participants will need to bring a laptop with R installed.
The aim of this two-day course is to enable students to carry out a variety of practical analyses using IRT in the R programming environment. We will work up to IRT models from Classical Test Theory (CTT) models comparing the advantages and disadvantages of each paradigm. Students will leave the course with the knowledge to begin using IRT to analyse their data in the R software package.
The course is delivered over 2 days, and focuses as an introduction on an overview of classical test theory and item response theory, and IRT model development (day 1), model fit, reliability, differential item functioning and test linking and equating (day 2). Our focus is on IRT models for dimension reduction and creation of latent variables (test scores), when we have categorical response data.
A mixture of lectures and computer practical sessions in R will be used to ensure participants appreciate the underlying statistical concepts and can apply the methods learned to real datasets containing either binary or polytomous (>2 category) outcomes. Participants will need to bring a laptop with R installed.
By the end of the course participants will have a good understanding of:
- philosophical issues surrounding the construction of scales and the delineation of constructs
- how CTT links to IRT and what advantages IRT brings to scale construction
- how to fit simple binary IRT models to data
- how to fit simple polytomous IRT models to data
- how to apply various methods of model fit assessment criteria in IRT
- the conceptualisation of reliability, as distinct but related to that in CTT, in the IRT framework of scale construction, and be able to estimate the reliability of a scale at a point on the continuum
- what causes items to function differently across different groups and be able to apply methods to find and deal with these items when they bias the results of the scale
- the issues surrounding linking the scores on two sets of people who take different tests measuring the same construct, and be able to apply some of the simpler linking techniques.
The day begins with an overview of the rationale and terminology used in psychometric research. It then outlines the underlying theoretical concepts and model specifications, focusing on logistic type models for binary response data and related extensions to polytomous response outcomes. Model development topics include the Rasch model, and the 1 parameter logistic (1PL), 2PL and 3PL IRT models, considering the underlying assumptions of each.
Day 2 will focus on how models are optimised and assessed for the type of response data used (model fit). Strategies are outlined to address the reliability of the scale. In particular we consider how to test for and handle differential item functioning. Day 2 concludes with practical guidance on different ways to approach test linking and equating when two similar but different tests (or questionnaires) exist that measure the same construct. Computer practicals are included on both days.
Early Bird rates apply if booked before February 28 2019
Early bird rate for Students £295 / Standard Rate £395
Early bird rate for Public Sector £395/ Standard Rate £495
Early bird rate for Commercial-Industry £495/Standard Rate £595
Please refer to the booking information for Keele internal rates.
Price includes refereshment's and lunches during the event.