ERIC ED563460: Evaluating a Strategy for Setting Cut Scores on a Computer Adaptive Test

2011·0.04 MB·English

by  ERIC

EEvalluattiing a SSttrattegy ffor
Setting Cut Scores on a
CComputter AAddapttiive TTestt
Deanna L. Morgan, The College Board
Chad W. Buckendahl, Alpine Testing Solutions
Introduction
Use of CATs have steadily increased
•
Currently planning underway for implementation of
•
Common Core Standards and the Smarter Balance
CCoonnssoorrttiiuumm iiss ffooccuusseedd aarroouunndd aa mmooddeell uussiinngg CCAATT ffoorr aa
large part of the assessment system.
SSeettttiinngg ccuutt ssccoorreess wwiitthh aa CCAATT ccaann bbee cchhaalllleennggiinngg
•
Potentially infinite number of test forms
•
AAddmmiinniissttrraattiioonn cchhaarraacctteerriissttiiccss mmaayy bbee uunnffaammiilliiaarr ttoo
•
panelists
“Wainer Method”
Modification of Angoff (1971) dichotomous
•
jjuuddggmmeenntt mmeetthhoodd tthhaatt lleevveerraaggeess aa CCAATT aallggoorriitthhmm
as experienced by the examinees.
SSttrattegy was concepttualliizedd bby LLunz ((22000000)) andd
•
Sireci and Clauser (2001) but limited empirical
eevviiddeennccee ooff ssuucccceessssffuull aapppplliiccaattiioonn iiss aavvaaiillaabbllee.
O’Neill, Tannenbaum, and Tiffen (2005) applied
•
tthhee mmeetthhooddoollooggyy ttoo TTOOEEFFLL iinn tthhee ccoonntteexxtt ooff tthhee
nursing licensure examination program.
Current Study
Applied “Wainer Method” to a fixed-length CAT as
•
oonnee ppaarrtt ooff aa ttwwoo ppaarrtt ssttuuddyy ffoorr aa mmuullttiippllee-cchhooiiccee
writing assessment. Only the part of the study
ppeerttaainingg ttoo tthee CCAT will bbee inccluuddeedd.
Three sources of validity evidence as described in
•
KKaannee ((11999944, 22000011)) wwiillll bbee uusseedd ttoo oorrggaanniizzee tthhee
presentation.
PPrroocceedduurraall, IInntteerrnnaall, EExxtteerrnnaall
••
Procedural
Panelists were solicited by the state higher
•
eedduuccaattiioonn ccoooorrddiinnaattiinngg bbooaarrdd ttoo rreepprreesseenntt
multiple stakeholder groups.
IInncclluuddeedd hhiigghh sscchhooooll tteeaacchheerrss, ddeevveellooppmmeennttaall wwrriittiinngg
•
instructors, freshman composition instructors, and
higher education administrators from across the
geographic regions and institution types in the state.
30 panelists – 26 females, 4 males
•
Teaching experience – Mean 18.2 years, SD 9.34 years
•
Procedural
Orientation including the purpose of the standard setting
•
meeting, intended use of the resultant cut score, and a
description of the test (including a high level overview of
the CAT algorithm) began the meeting.
PPanelliistts were bbrokken iintto 55 smallller groups off 66 peoplle tto
•
discuss and draft performance level descriptors (PLDs) of
the borderline examinee.
Large group discussion of the drafts created in the small
•
groups followed and was led by the facilitator to produce
the final PLDs which were then transcribed, copied, and
distributed.
Procedural
Panelists received training on the methodology
•
wwhheerree tthheeyy wweerree iinnssttrruucctteedd ttoo ttaakkee tthhee tteesstt
responding to each question either correctly or
inccoorreeccttlyy aass ttheeyy woouuldd eexppeecctt tthee bboorddeerlinee
examinee to respond without consideration for
which incorrect answer to choose – only that it
was incorrect.
Duringg trainingg ppanelists took the test multipple
•
times changing their response patterns to help
them understand the adaptive nature of the test
Procedural
Panelists completed two rounds of ratings by
•
ttaakkiinngg tthhee tteesstt aass iiff tthheeyy wweerree tthhee bboorrddeerrlliinnee
examinee.
BBettween RRoundd 11 andd RRoundd 22 off rattiings, a llarge
•
group discussion was held and feedback provided
wwhhiicchh iinncclluuddeedd tthhee iinnddiivviidduuaall ccuutt ssccoorree ooff eeaacchh
panelist, the panel’s average and median cut
score,, and imppact data based on a nationallyy
representative norm group of examinees who
have taken the test.
Procedural
Panelists completed an evaluation form following Round 2
•
concerning:
Efficacy of the orientation
•
Understanding of the PLDs and borderline examinee
•
Training on the rating task
•
Helpfulness of discussion and feedback
•
LLevell off conffiiddence iin tthhe resullttiing sttanddardds
•
All mean ratings for training and adequacy of time allowed
•
wweerree bbeettwweeeenn 33.88 aanndd 44.00 oonn aa ssccaallee ffrroomm 11==
Unsuccessful/Inadequate to 4 = Successful/Adequate.
CCoonnffiiddeennccee rraattiinnggss rraannggeedd ffrroomm 11 = NNoott CCoonnffiiddeenntt ttoo 44 =
•
Confident with a mean rating of 2.75.
Procedural
Panelist comments on the evaluation forms and during
•
Round 2 discussion indicated that there was some
confusion about the rating task during Round 1.
Specifically, panelists spent a large amount of time
•
ddebbattiing over whhiichh iincorrectt answer tto chhoose even
though instructed that it did not matter which as long as it
was incorrect when apppproppriate.
Additionally, one panelist indicated that they did not
•
understand during Round 1 that they were taking it as the
borderline student and had an “Aha!” moment during the
discussion between rounds of ratings.