Generalizability Theory: R Statistics

This performs the specified G-Theory analysis using the R Statistics package "gtheory"

 

Instructions for Brennan.3.2.txt G-Theory Analysis

Instructions for Rajaratnam.2.txt G-Theory Analysis

 

 

 

Define G-Theory Model

a Generalizability G-Study and a Decision D-study are specified and performed

Measurement facets; (or factors)

(any facet. Default is all facets.)

the facets used to explain the variance in the data

Select interactions

(any pair of facets. Default is none,()

the facet interactions  used to explain the variance in the data. Can specify facets not checked as Measurement facets.

Score variable. (Default is Observation)

(one of Obs, Stp, Exp, Res, Var, StRes, Wt, LProb, Meas, Disp, MPCat)

the raw data in the Residual file whose variance is to be explained. Examples: the observations, the standardized residuals. See Residual File

Object of Measurement

(any facet. Default is first facet.)

the facet whose Generalizability (Reliability) is to be estimated. This is automatically added to the Measurement facets, if not checked there.

Strata facet

(any facet. Default is none.)

facets which defines strata in the data, such as subtests of items, content strands. Can specify a facet not checked as a Measurement facet.

 

R Statistics package "gtheory", described in cran.r-project.org/web/packages/gtheory/gtheory.pdf, is launched from this window. It is based on Brennan, R. L. (2001). Generalizability theory. New York: Springer, and Rajaratnam, N., Cronbach, L. J., & Gleser, G. C. (1965). Generalizability of stratified-parallel tests. Psychometrika, 30(1), 39-56. It includes example datasets: data(Brennan.3.2) and data(Rajaratnam.2). These are slightly reformulated as Facets example datasets: Brennan.3.2.txt and Rajaratnam.2.txt. No changes are made to the observations, but the element numbers differ resulting in some differences in the output of "gtheory".

 

Example: "gtheory" output for Brennan.3.2.txt in the Examples folder:

 

Brennan.3.2.txt Facets specifications:

Title= G-Theory Brennan.3.2

Facets = 3

Models = ?,?,?,R9

Positive = 1,2,3

Noncenter=2

Labels=

1, Task

1-3

*

2, Person

1-10

*

3, Rater, G ; group anchoring: each rater rated one task

1_4, Group 1, 0, 1

5-8, Group 2, 0, 2

9-12, Group 3, 0, 3

*

 

data=

;    Task Person Rater Score

1      1     1     5

1      2     1     9

................

 

With 2 facets: (omitting tasks):

    source       var percent n

1    Rater 0.8840014    20.6 1

2   Person 0.6257537    14.6 1

3 Residual 2.7872078    64.9 1

 

$generalizability

[1] 0.7292988

 

With 3 facets (raters are nested within task):

    source       var percent n

1    Rater 0.6068375    13.8 1

2   Person 0.6257565    14.2 1

3     Task 0.3811112     8.7 1

4 Residual 2.7872049    63.3 1

 

$generalizability

[1] 0.7292999

 

With 3 facets and interactions:

       source       var percent n

1 Task:Person 0.5595713    12.8 1

2  Task:Rater 0.2952631     6.7 1

3       Rater 0.3522727     8.0 1

4      Person 0.4731462    10.8 1

5        Task 0.3251216     7.4 1

6    Residual 2.3802471    54.3 1

 

$generalizability

[1] 0.5514371


Help for Facets (64-bit) Rasch Measurement and Rasch Analysis Software: www.winsteps.com Author: John Michael Linacre.