MEASUREMENT RESEARCH ASSOCIATES
TEST INSIGHTS
February 2010
Greetings,
 

Examiner consistency within his/her own rating pattern in an oral examination should be monitored.  This brief study explores whether an examiner's internal consistency is related to his/her severity. 


Lidia Martinez
Manager, Test Development and Analysis

The Relationship between Examiner Severity and Consistency
Examiner severity is the convenient term for the tendency of an examiner to give lower ratings or higher ratings. This tendency towards severity or leniency is due to examiner expectations, characteristics, and standards. A severity measure for each examiner in the many-facet Rasch analysis is calculated using all of the ratings the examiner gave during the course of the examination.

Examiner consistency is measured by a mean-square fit statistic. This statistic is based on the ratio of observed error variance to expected error variance.  It's expected value is 1 (i.e., a ratio of 1:1). The mean square fit statistic for an examiner indicates his/her consistency or how well his/her pattern of ratings meet expectations given examiner severity and candidate ability (i.e., fit to the model). Neither too high nor too low fit statistics are desirable. 

When the examiner's fit statistic is less than .5, it indicates over 50% less variance in his/her ratings than is expected. It is likely that the examiner tended to give many candidates the same rating, regardless of their ability. This type of examiner is not only too predictable, but he/she is not distinguishing differences among candidates. When the fit statistic is greater than 1.5, it indicates over 50% more variance in his/her ratings than is expected. It is likely that the examiner gave candidates unexpectedly high or low ratings compared to their overall ability.

The question is whether there is a correlation between measured examiner severity and examiner consistency (outfit mean square fit statistic). To study this question, random performance examinations were selected and the Pearson correlation between severity and consistency for the examiners was calculated.

The table below shows that there are low, non-significant correlations between examiner severity and consistency.  The table also shows that the vast majority of the examiners meet the criteria for consistency.  The low correlations between severity and consistency show that 1) most examiners are internally consistent in their rating of candidates; 2) that examiners, regardless of their measured severity, tend to be consistent in their rating of candidates; and 3) that severity does not predict consistency or vice-versa.  The low numbers of inconsistent examiners reflects good examiner training and an understanding of the rating process.

 

Exam

N of Examiners

Correlation between Severity and Consistency

Significance

(ns = not significant)

Number  (%) of

inconsistent examiners

Exam 1

44

.05

ns

0

Exam 2

24

-.14

ns

0

Exam 3

72

.00

ns

4 (5%)

Exam 4

146

-.09

ns

4 (3%)

Exam 5

81

-.06

ns

2 (2%)

 


Measurement Research Associates, Inc.
505 North Lake Shore Dr., Suite 1304
Chicago, IL  60611
Phone: (312) 822-9648     Fax: (312) 822-9650


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