+ All Categories
Home > Science > Waop presentation 2016 a basket case

Waop presentation 2016 a basket case

Date post: 22-Jan-2018
Category:
Upload: jelle-geertsma
View: 29 times
Download: 2 times
Share this document with a friend
12
eelloo.nl A basket case Using baskets and verification in unproctored testing Jozien Silvester Liesbeth van Leeuwen Jelle Geertsma WAOP Presentation 25 november 2016
Transcript
Page 1: Waop presentation 2016 a basket case

eelloo.nl

A basket caseUsing baskets and verification in unproctored testing

Jozien SilvesterLiesbeth van LeeuwenJelle Geertsma

WAOP Presentation 25 november 2016

Page 2: Waop presentation 2016 a basket case

eelloo.nl

Unproctored testing

• Stakes are high in selection

• Possibility of cheating

• Proctoring not always possible or efficient

• Benefits for candidates of unproctored testing: time and place of choice

Page 3: Waop presentation 2016 a basket case

eelloo.nl

Unproctored testing

• Validity proctored /unproctored equal if:

• Informing candidates about optimal test conditions

• Taking measures to prevent fraud

• Items that measure reasoning (fluid)

• Time limit

• Basket method

Page 4: Waop presentation 2016 a basket case

eelloo.nl

Basket method

• Large item bank

• Items of equal difficulty are placed in a basket

• Item is randomly picked from basket

• Same number of baskets for every candidate

• Four subtests: Figure series, Analogies, Diagrams, Number series

Page 5: Waop presentation 2016 a basket case

eelloo.nl

Benefits of basket method

• Easy calculation

• Easy monitoring, maintaining, extension

• Easy retesting (system remembers which items were already administered)

• Prevention of item leakage

Page 6: Waop presentation 2016 a basket case

eelloo.nl

Validity

• High correlation with proctored extended test (.75)

• Test-retest correlation (.77)

• About 8% scores lower than expected, possible cheaters

Page 7: Waop presentation 2016 a basket case

eelloo.nl

Basket monitoring

Page 8: Waop presentation 2016 a basket case

eelloo.nl

Basket monitoring

• Some items with equal difficulty are more equal than others

Basket Item Difficulty (n) Difference Average

1017 Hcyf106b 0.66 (287) 0.10 0.68

Hcyf106c 0.65 (266)

HCYF01b 0.74 (294)

1018 HCYF112 0.62 (409) 0.02 0.63

HCYF112b 0.64 (438)

Page 9: Waop presentation 2016 a basket case

eelloo.nl

Difference in p-value

• Even if items are carefully balanced, there will still bedifferences, although they may be small

• For most candidates this will average

• Some will have all the ‘easy’ or all the ‘difficult’ items

• Effect on reliability and score

• Comparison of correction methods on two measures

Page 10: Waop presentation 2016 a basket case

eelloo.nl

Score calculation improvement - reliabilities (split half)

Education level N CTT IRT1pl IRT2pl Elo Glicko

MBO (Vocational) 743 0.85 0.80 0.81 0.79 0.86

HBO (Bachelor) 1397 0.82 0.82 0.83 0.79 0.85

WO (Master) 706 0.80 0.76 0.77 0.74 0.81

• Glicko has best performance

• CTT has good results as well

Page 11: Waop presentation 2016 a basket case

eelloo.nl

Effects Glicko on z-score

Item difficulty group

regular scoreGlicko score

z-sc

ore

Page 12: Waop presentation 2016 a basket case

eelloo.nl

Concluding remarks

• Valid unproctored test possible with basket method

• Glicko shows slight improvement in score calculation, but CTT also works well


Recommended