How to - Chiang Mai University
Download
Report
Transcript How to - Chiang Mai University
MEQ
Analysis
Outline
Validity
Reliability
Difficulty Index
Power of Discrimination
Validity
“evidence present to support or
refute the meaning assigned to
assessment results”
face validity
content validity
criterion-related validity
construct validity
Face Validity
high face validity
knee jerk & nervous system
seemed valid (only)
~ person
for doctor => high face validity
for lay people => low face validity
MEQ => high face validity
Content Validity
~ sample & population
MCQ => high content
validity
MEQ => low content
validity
Validity
“evidence present to support or
refute the meaning assigned to
assessment results”
- don’t need the score
face validity
- before using the
test
content validity
criterion-related validity - score needed
- after using
construct validity
the test
Criterion-Related
Validity
Predictive validity
MEQ score & close observation
score of real practice at ER
Concurrent Validity
MEQ score & VIVA score in the
same topic
Statistic = correlation coefficient
Construct Validity
based on theory
communication skill ~
leadership
correlation OSCE : communication
VS questionnaire : leadership
good ethics ~ beloved doctor
MEQ : medical ethics VS
questionnaire : beloved doctor?
Reliability
stability
internal consistency
equivalent
Stability
test-retest reliability
parallel form reliability
intra-rater reliability
~ scoring key
statistics : correlation coeff.
0-1
Internal Consistency
[homogeneity]
item - item correlation
item - total correlation
split half correlation
Item-Item Correlation
each item
Dichotomous
Interval
Phi Correlation
Pearson’s Product Moment
Correlation
whole test
Mean of ...
Item-Item Correlation
Item-Total Correlation
each item
Dichotomous
Interval
Point Biserial Correlation
Pearson’s Product Moment
Correlation
whole test
Mean of ...
Item-Total Correlation
Spilt Half Reliability
Dichotomous
Kuder-Richardson 20 (KR 20)
Interval
Kuder-Richardson 21 (KR 21)
Cronbach’s alpha coefficient
Equivalent
parallel item on alternate
form reliability
inter-rater reliability
agreement
kappa
Difficulty Index
[p]
(mean H + mean L)/2(full score)
p = 1 => very easy
p = 0 => very difficult
expecting p = 0.2-0.8
must => 0.7
should => 0.5
Power of Discrimination
[r]
(mean H - mean L)/(full score)
> 0.40 => very good item
0.30 - 0.39 => good item
0.20 - 0.29 => borderline
< 0.20 => poor item
Conclusion
Validity
Reliability
Difficulty Index
Power of Discrimination