For short tests of i=1,L dichotomous items taken by n=1,N subjects, we can easily obtain conditional-maximum-likelihood estimates of the item difficulty parameters, Di, of the Rasch model using the logit fixed-effects estimator in the clogit and xtlogit, fe commands of Stata. Data may be unbalanced.
These commands require that all responses (usually coded, 0/1) are stored in separate observations, while a group-variable is used to identify the observations that belong to the same subject. The reshape command may be useful to convert data into this format.
We describe the Rasch model to Stata as a "conditional logit model" with L covariates x(nik), k=1,L, so that x(nik) = -1 if i=k, and 0 otherwise. The regression coefficient of x(..k) becomes Dk. In order to set the origin of the Rasch measurement scale, Stata will automatically withdraw the first predictor variable D1 and fix it at zero. Alternatively, and better, is to withdraw the most stable of the Di yourself.
|Conditional (fixed-effects) logistic regression|
|Number of observations = 1050|
Here is the procedure for obtaining item calibrations:
(1) if necessary, transform the data into long-format
reshape item*, n(subject_id) j(ItemID)
(2) create explanatory variables for the Di
for num 1/L : gen D = -(X==ItemID)
(3) estimate the Di parameters
clogit item D*, group(subject_id)
xtlogit item D*, n(subject_id) fe
|Subject Score Distribution|
|dropped as extreme|
dropped as extreme
Example: The responses of 120 subjects to 10 math problems are coded as 1 (correct) and 0 (incorrect) - see Subject Score Distribution Table. I want to know whether the 10 problems maintain their difficulties, regardless of the subjects' abilities.
Apply clogit - the box above shows selected output.
To examine item difficulty invariance, we compare the item calibrations obtained from the full sample with those obtained from the low-performing (scores<=5) and high-performing (scores>=5) subjects. The box below shows selected results.
We are pleased to see that for low scorers, the items are exhibiting statistical invariance according to the hausman test. Nevertheless, items 7, 9 and 10 (the harder items) are somewhat problematic, meriting further investigation. For high scorers, it appears that almost all items have become about 1 logit more difficult. In fact, it is item 1 that has become about 1 logit easier. Item 1 is the easiest item and is most off target for the high scorers. This illustrates the need to choose the fixed (i.e., withdrawn) item carefully in order to avoid confusion in interpreting results. After an adjustment is made for item 1, the item difficulties are seen to be statistically invariant also for the high scorers.
If the wording of the items is known, then producing an item map of the construct based on these item difficulty calibrations is straightforward via the Stata graphical plotting command graph.
Once item difficulties have been computed, the subject measures may be estimated (via maximum-likelihood, conditional on the estimates of the item parameter) via logit (combined with byvar).
With the Stata command xtlogit, re, it is easy to estimate the Rasch model in which the person-parameters are treated as random effects, conditional on person-level variables.
For complete details of this analysis, see www.stata.com/support/faqs/stat/rasch.html
Department of Sociology/ICS
Estimating Rasch Models with Stata: Conditional (fixed-effects) Logistic Regression, Weesie J. Rasch Measurement Transactions, 2000, 13:4 p. 724
|Rasch Measurement Transactions (free, online)||Rasch Measurement research papers (free, online)||Probabilistic Models for Some Intelligence and Attainment Tests, Georg Rasch||Applying the Rasch Model 3rd. Ed., Bond & Fox||Best Test Design, Wright & Stone|
|Rating Scale Analysis, Wright & Masters||Introduction to Rasch Measurement, E. Smith & R. Smith||Introduction to Many-Facet Rasch Measurement, Thomas Eckes||Invariant Measurement: Using Rasch Models in the Social, Behavioral, and Health Sciences, George Engelhard, Jr.||Statistical Analyses for Language Testers, Rita Green|
|Rasch Models: Foundations, Recent Developments, and Applications, Fischer & Molenaar||Journal of Applied Measurement||Rasch models for measurement, David Andrich||Constructing Measures, Mark Wilson||Rasch Analysis in the Human Sciences, Boone, Stave, Yale|
|in Spanish:||Análisis de Rasch para todos, Agustín Tristán||Mediciones, Posicionamientos y Diagnósticos Competitivos, Juan Ramón Oreja Rodríguez|
|Forum||Rasch Measurement Forum to discuss any Rasch-related topic|
Go to Top of Page
Go to index of all Rasch Measurement Transactions
AERA members: Join the Rasch Measurement SIG and receive the printed version of RMT
Some back issues of RMT are available as bound volumes
Subscribe to Journal of Applied Measurement
Go to Institute for Objective Measurement Home Page. The Rasch Measurement SIG (AERA) thanks the Institute for Objective Measurement for inviting the publication of Rasch Measurement Transactions on the Institute's website, www.rasch.org.
|Coming Rasch-related Events|
|June 23 - July 21, 2023, Fri.-Fri.||On-line workshop: Practical Rasch Measurement - Further Topics (E. Smith, Winsteps), www.statistics.com|
|Aug. 11 - Sept. 8, 2023, Fri.-Fri.||On-line workshop: Many-Facet Rasch Measurement (E. Smith, Facets), www.statistics.com|
The URL of this page is www.rasch.org/rmt/rmt134g.htm