Due to the high-stakes nature of
large-scale assessment, it is sensible to ensure that results from the
assessments are based on effective instruction and true student achievement.
U.S. Secretary of Education Arne Duncan issued a policy letter, dated June 24,
2011, that urges states to “make assessment security a high priority” and to
“ensure that assessment development contracts include support for activities
related to test security, including forensic analysis.” Additionally, it is
recommended by the Association of Test Publishers and the Council of Chief
State School Officers (2010) that rules and procedures be adopted that respond
to instances of test administration irregularities.
To assist in meeting these
recommendations, DRC protects its clients’ investments in testing through the
use of its industry-leading research on erasure behavior, expert knowledge of
the tests’ underlying measurement models, and other statistical indicators of
DRC’s expertise means clients have access to the industry’s best
tools to continuously protect the integrity of their assessment and
Scanning technology used by DRC provides
detailed erasure analysis at the student level. This analysis takes data from
the multiple-choice item responses of all students and analyzes it at the state,
school, and student levels for evidence of responses being erased and changed.
Erasures are not made when students are taking a computer-based assessment. Answer changes, however, are still made.
DRC's online test engine, DRC INSIGHT,
is able to track student progress throughout the test administration. If a
student changes a multiple-choice answer, the system is able to track the
answer that was initially given, what answer was finally selected, and
everything in between.
Statistical indices are used to identify potential testing
irregularities and separate meaningful gains from the spurious.
Use of Multiple Measures
DRC uses a variety of traditional and
advanced statistical analyses to detect aberrances in test-taking behavior.
These techniques provide for improved prediction and increased sensitivity. The
methods used are as follows: