Validity, Invariant Measurement, and Rater-Mediated Assessments

Author(s):  
George Engelhard ◽  
Stefanie A. Wind
2018 ◽  
Vol 11 (3) ◽  
pp. 205979911881439 ◽  
Author(s):  
Stefanie A Wind

Model-data fit indices for raters provide insight into the degree to which raters demonstrate psychometric properties defined as useful within a measurement framework. Fit statistics for raters are particularly relevant within frameworks based on invariant measurement, such as Rasch measurement theory and Mokken scale analysis. A simple approach to examining invariance is to examine assessment data for evidence of Guttman errors. I used real and simulated data to illustrate and explore a nonparametric procedure for evaluating rater errors based on Guttman errors and to examine the alignment between Guttman errors and other indices of rater fit. The results suggested that researchers and practitioners can use summaries of Guttman errors to identify raters who exhibit misfit. Furthermore, results from the comparisons between summaries of Guttman errors and parametric fit statistics suggested that both approaches detect similar problematic measurement characteristics. Specifically, raters who exhibit many Guttman errors tended to have higher-than-expected Outfit MSE statistics and lower-than-expected estimated slope statistics. I discuss implications of these results as they relate to research and practice for rater-mediated assessments.


Author(s):  
Antonio Pifferi ◽  
Davide Contini ◽  
Lorenzo Spinelli ◽  
Alessandro Torricelli ◽  
Rinaldo Cubeddu ◽  
...  

1983 ◽  
Vol 14 (1) ◽  
pp. 19-32 ◽  
Author(s):  
Zekâi Şen

A prediction model capable of aggregation with recursive unbiased minimum variance estimation algorithms based on the Kalman filter technique has been formulated and applied for predicting monthly flows such that their summation is equal to annual flow in the same year. The model represents a discrete linear stochastic system where the states are defined as monthly flows in addition to the measurement equation with time invariant measurement matrix and annual flow measurement. Provided that observed or generated annual flows are available then the proposed model can be employed to predict monthly flows so that their aggregation yields the total annual flow.


Sign in / Sign up

Export Citation Format

Share Document