Rasch Factor Analysis

Conventional Factor Analysis input xni is a test score, Likert rating or MCQ response of person n to item i. Incommensurable origins and scales are reconciled by using sample means and standard deviations to commensurate items:
(1)

This "equating" depends on complete data. When data are missing, they must be feigned or incomplete persons or items deleted. The model for Factor 1 is:
(2)

{un1} n=1,N is a vector of person "factor scores"
{vi1} i=1,L is a vector of item "factor loadings"

The Factor 1 residual is:
(3)

Whether this residual is all error or contains other factors is unknown. To find out, the residual is taken as the data for Factor 2. Matrices of decreasing residuals {{znij}} j=2,M are extracted in turn to calculate the M factor model:
(4)

Optimal values for person and item vectors {unj} and {vij} for each Factor j in turn are found by minimizing:
(5)

Decomposition to identify each Factor j is done by initializing at wnj=1 and iterating equations:
(6)

until changes in {wnj} become uninteresting.

Factor score wnj is the value predicted by Factor j's regression on "independent" variables i = 1,L with regression coefficients {vij}.

Problems with Factor Analysis of Observational Data:

1. Since, data {{xni}} are never linear measures, Equations (2) and (3) cannot work in a reproducible way.

2. The necessity for complete data is awkward.

3. The residuals {{znij}} for seeking smaller factors are awash in turbulence left by preceding larger factors.

4. There is no way to know when to stop factoring.

5. Few algorithms provide standard errors for factor loadings or factor scores.

6. When old items are refactored from new data, factor sizes, loadings and number of factors change.

Most factor analysts can tolerate problems 1-5. Problem 6, however, is impossible to swallow. As analysts notice the numerical fragility of their structures, they abandon their person factor scores and ignore everything in the factor loadings except which factor gives each item its highest loading.

Person scores for these "same factor" clusters are obtained, not from factor scores, but by summing the standardized person responses zni1 to the clustered items.

Rasch Factor Analysis:

Since data {{xni}} are labels for ordinally interpreted qualities, we address them with a probability model for ordered categories. For a dichotomy, we recognize xni=0,1 as a binomial. The error model which follows is not the ill-chosen linear error model of factor analysis, but a binomial probability Pnix for the occurrence of xni:
(7)

from which

Parameters, as with factor analysis (5), can be estimated by minimizing:
(8)

but now we get, not only least-square (maximum likelihood) estimates of linear person measures Bn and linear item calibrations Di on their common variable, but also a stochastic basis, the binomial error variances pni1pni0, for estimating useful standard errors and evaluating the probabilities of residuals:
(9)

This enables detailed misfit analysis which, in turn, allows a partition of the matrix of residuals {{zni1}} into those many zni1 which are observed to be no greater than the probability model expects, say |zni1| < 2, and the subset of more extreme residuals, say |zni1| > 2 or 3, which are sufficiently improbable to invite investigation.

It is only when improbable residuals emerge that there is an incentive to look for a second variable. The improbable residuals tell us where to look. Should there be another useful variable in these data, it will be most evident among the responses to the subset of items and persons which misfit the first analysis.

To seek a second variable, therefore, we apply the probability model again, not to the whole matrix of residuals {{zni1}}, but only to the original ordinal responses {{xni}} to this subset of items. We estimate a new set of linear item calibrations {Di2} for these items, along with a new set of additive conjoint person measures {Bn2} on this newly defined "Variable 2". To find out whether Variable 2 is useful, we plot person measures {Bn2} against person measures {Bn1}. This plot will show us the extent to which we have found a useful second variable.

Benjamin D. Wright 1994 RMT 8:1 p. 348-9


Rasch factor analysis Wright BD. … Rasch Measurement Transactions, 1994, 8:1 p.348-9



Rasch Publications
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

To be emailed about new material on www.rasch.org
please enter your email address here:

I want to Subscribe: & click below
I want to Unsubscribe: & click below

Please set your SPAM filter to accept emails from Rasch.org

www.rasch.org welcomes your comments:

Your email address (if you want us to reply):

 

ForumRasch 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
May 17 - June 21, 2024, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com
June 12 - 14, 2024, Wed.-Fri. 1st Scandinavian Applied Measurement Conference, Kristianstad University, Kristianstad, Sweden http://www.hkr.se/samc2024
June 21 - July 19, 2024, Fri.-Fri. On-line workshop: Rasch Measurement - Further Topics (E. Smith, Winsteps), www.statistics.com
Aug. 5 - Aug. 6, 2024, Fri.-Fri. 2024 Inaugural Conference of the Society for the Study of Measurement (Berkeley, CA), Call for Proposals
Aug. 9 - Sept. 6, 2024, Fri.-Fri. On-line workshop: Many-Facet Rasch Measurement (E. Smith, Facets), www.statistics.com
Oct. 4 - Nov. 8, 2024, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com
Jan. 17 - Feb. 21, 2025, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com
May 16 - June 20, 2025, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com
June 20 - July 18, 2025, Fri.-Fri. On-line workshop: Rasch Measurement - Further Topics (E. Smith, Facets), www.statistics.com
Oct. 3 - Nov. 7, 2025, Fri.-Fri. On-line workshop: Rasch Measurement - Core Topics (E. Smith, Winsteps), www.statistics.com

 

The URL of this page is www.rasch.org/rmt/rmt81q.htm

Website: www.rasch.org/rmt/contents.htm