basic syntax: rwg var, scale(#) model(#)
Description: Rwg calculates single(model 1) and multiple items(model 2) within group interra > ter reliability. The program requires data in "long" format with panel setup > (tsset). The program makes use of panel ids to identify raters and ratees, th > erefore, it is very important to correctly specify items and judges in panel > setup.
Multiple-items estimator (model 2) and 5 points response scale are set as defau > lt.
*Model(1): Single-item estimator
Rwg(i)= 1 - [Sx^2/seu]
*Model(2): Multiple-item estimator
J[1-(Sx^2/seu)] Rwg(j)= --------------------------- J[1-(Sx^2/seu)]+ (Sx^2/seu)
Where Rwg(i) and Rwg(j) are within-group interrater reliability for single and > multiple items. Sx^2 is the observed variance of ratings, J is number of item > s (model 2 eq.), seu=(A^2-1)/12 and `A' is response scale(e.g. 5 or 7 points > likert scale).
Examples: ---------- tsset items judges
rwg x, scale(7) model(1) rwg x, scale(9) model(1)
rwg x, scale(7) model(2) rwg x, scale(9) model(2)
Author Muhammad Rashid Ansari INSEAD Business School 1 Ayer Rajah Avenue, Singapore 138676 > rashid.ansari@insead.edu
References: James, L. R., Demaree, R. G., & Wolf, G. 1993. rwg: An assessment of within-gro > up interrater agreement. Journal of Applied Psychology, 78: 306-309.
James, L. R., Demaree, R. G., & Wolf, G. 1984. Estimating within-group interrat > er reliability with and without response bias. Journal of Applied Psychology, > 69, 85-98
Finn, R. H (1970) A note on estimating the reliability of categorical data. Edu > cational and Psychological Measurement, 30, 71-76