16.00 o6.2 d lillis

36
STATISTICAL MODELLING FOR IMPROVEMENT OF SECONDARY AND TERTIARY PHYSICS ASSESSMENTS

Upload: nzip

Post on 28-Nov-2014

376 views

Category:

Technology


2 download

DESCRIPTION

Physikos 2: David Lillis

TRANSCRIPT

Page 1: 16.00 o6.2 d lillis

STATISTICAL MODELLING

FOR IMPROVEMENT OF SECONDARY AND TERTIARY

PHYSICS ASSESSMENTS

Page 2: 16.00 o6.2 d lillis

PHYSICS: EXTERNAL RESULTS LEVEL 3 BY GENDER (2010)

NUM N A M E

Male 14,630 29.4 47.8 15.9 6.8

Female 8,375 29.5 50.6 14.5 5.4

Page 3: 16.00 o6.2 d lillis

LEVEL 3 EXTERNAL STANDARDS (2010 RESULTS)

STD TITLE NUM N A M E

90520 Demonstrate understanding of wave systems 6,443 27.2 54.6 13.3 4.9

90521 Demonstrate understanding of mechanical systems

6,480 30.7 46.6 16.6 6.1

90522 Demonstrate understanding of atoms, photons and nuclei

4,862 32.4 46.2 15.1 6.4

90523 Demonstrate understanding of electrical systems

5,220 27.9 47.0 16.9 8.2

Page 4: 16.00 o6.2 d lillis

PROFILES OF EXPECTED PERFORMANCE (PEPs)

Expected percentage band for each grade, based on

statistical information & changes in examination

format etc

Provide guidance to examiners & markers

Ensure results that are consistent with the standards

and that standards are maintained

Page 5: 16.00 o6.2 d lillis

2010 PEP BANDS ACTUAL: N = 32.3%; A = 46.2%; M = 15.1%; E = 6.4%

Page 6: 16.00 o6.2 d lillis

COHORT STRENGTH & STANDARD DIFFICULTY

Sk = Σj ρkjnkj [ Rkj(j) – Rj(j) ] / Σj nkj

Dk = Σj ρkjnkj [ Rkj(j) – Rkj(k) ] / Σj nkj

Page 7: 16.00 o6.2 d lillis

RESULTS: PHYSICS 90522

Q1ai Q1aii Q1b Q1c Q1d Q2a Q2b Q2c Q2d Q3a Q3b Q3c FINAL

1 -99 2 1 2 3 2 2 1 3 1 2 2

1 1 2 3 1 1 2 1 1 3 2 1 2

2 3 1 3 1 -99 -99 -99 -99 -99 -99 -99 1

1 3 1 3 4 3 3 2 4 3 1 2 3

2 2 1 3 4 3 2 1 4 2 1 -99 2

2 3 2 -99 2 3 3 2 4 3 3 -99 3

Page 8: 16.00 o6.2 d lillis

PHYSICS 90522 Principal Components Analysis

Page 9: 16.00 o6.2 d lillis

PHYSICS 90254: Waves (Level 2)

Page 10: 16.00 o6.2 d lillis

ECONOMICS 90631 (Level 3)Market failure and government interventions

Page 11: 16.00 o6.2 d lillis

CHEMISTRY 90696 (Level 3) Oxidation-reduction processes

Page 12: 16.00 o6.2 d lillis

STATISTICS 90643 (Level 3)Probability

Page 13: 16.00 o6.2 d lillis

PHYSICS 90522: ITEM LOADINGS THE FIRST & SECOND DIMENSIONS

ITEM

DIM 1

DIM 2

Q1ai 0.27 0.29

Q1aii 0.33 0.18

Q1b 0.14 0.21

Q1c 0.31 -0.28

Q1d 0.31 -0.26

Q2a 0.36 -0.06

Q2b 0.28 0.4

Q2c 0.19 -0.58

Q2d 0.35 -0.25

Q3a 0.24 0.04

Q3b 0.27 0.35

Q3c 0.34 -0.01

Page 14: 16.00 o6.2 d lillis

INTERPRETATION?

FIRST DIMENSION

Understanding of atoms, photons and nuclei

SECOND DIMENSION

Item type (qualitative or quantitative responses)

Page 15: 16.00 o6.2 d lillis

MEASURES OF INTERNAL CONSISTENCY

Inter-item Correlation, Item Total Correlation &

Cronbach Alpha

Range between – 1 and + 1

Optimal between about 0.4 and 0.85

Page 16: 16.00 o6.2 d lillis

INTER-ITEM & ITEM TOTAL CORRELATIONS (CRONBACH ALPHA = 0.76)

ITEM Q1 Q2 Q3 TOTAL

Q1 1.00 0.50 0.48 0.58

Q2 0.50 1.00 0.45 0.57

Q3 0.48 0.45 1.00 0.58

Page 17: 16.00 o6.2 d lillis

EXAMINATIONS IN PHYSICS . . .

Always assess one dominant dimension (cognitive

construct) and at least one other dimension

For practical purposes, form a psychometric scale,

but not a perfect scale

Page 18: 16.00 o6.2 d lillis

ITEM RESPONSE THEORY

The probability of obtaining in a single item a particular grade or better (Not Achieved, Achieved, Merit or Excellence), for a candidate of ability θ (measured over the entire assessment), is given by the expression:

Pj = { 1 + exp [ ka ( bj – θ ) ] } -1

j indexes the assessment grades Achieved (A) or better, Merit (M) or better, and Excellence (E)

θ is the calculated ability (which you can also think of as a measure of performance)

a is the item discrimination bj is the estimated difficulty of gaining either an A or better, M or better,

or an E grade for the item. k = -1.7 (scales the logistic curve to approximate a cumulative ogive)

Page 19: 16.00 o6.2 d lillis

ITEM-CHARACTERISTIC CURVE (DICHOTOMOUS)DIFFICULTY: the ability at which P = 0.5

DISCRIMINATION: slope at that point

Page 20: 16.00 o6.2 d lillis

ITEM-CHARACTERISTIC CURVE (POLYTOMOUS)

Page 21: 16.00 o6.2 d lillis

IRT PARAMETERS

Item discriminations & grade difficulties estimated

using various models

Ability estimated from performance on all items,

taking account of item difficulties & discriminations

The most objective measure of performance?

Page 22: 16.00 o6.2 d lillis

90522: EXAMPLE ITEM PARAMETERS (Graded Response Model)

ITEM Discrimination Difficulty (AME)

Difficulty (ME)

Difficulty (E)

Q1aii 1.45 -1.03 0.44 2.08

Q2a 1.70 -1.01 -0.65 4.66

Page 23: 16.00 o6.2 d lillis

PHYSICS 90522: HISTOGRAM OF CANDIDATE ABILITIES (PERFORMANCES)

Page 24: 16.00 o6.2 d lillis

PHYSICS 90522: ITEM CHARACTERISTIC CURVES

Page 25: 16.00 o6.2 d lillis
Page 26: 16.00 o6.2 d lillis
Page 27: 16.00 o6.2 d lillis

THE SUFFICIENCY METHOD OF ASSIGNING OVERALL GRADES

For example:

5A => Achieved

3As + 3Ms => Merit

2Ms + 2Es => Excellence

Page 28: 16.00 o6.2 d lillis

ALIGNMENT OF ACTUAL FINAL GRADES WITH ABILITY-BASED GRADES

Actual Grade

N A M E Candidates

N 78.6 21.4 0.0 0.0 173

Theta A 10.2 79.0 10.2 0.6 353

Grade M 0.8 29.3 58.5 11.4 123

E 0.0 2.0 30.0 68.0 50

Page 29: 16.00 o6.2 d lillis

THE SCORE-GRADE METHOD

N A M E

0 1 2 3 4 5 6 7 8

Page 30: 16.00 o6.2 d lillis

SCORE-BASED MARKING

Improved discrimination between grades

Fairer outcomes

Reduction in year-to-year variability

Page 31: 16.00 o6.2 d lillis

NEW ZEALAND SCHOLARSHIP

SUBJECT Y 13 Cohort

Entries (male)

Entries (female)

Valid results

N S O

Physics 6938 961 323 1066 858 178 24

Chemistry 7323 962 650 1331 1119 188 24

Calculus 7511 1100 435 1263 1026 201 30

Statistics 14466 1128 619 1418 975 395 45

English 13561 683 1042 1339 919 362 53

ALL 9280 9676 14403 10995 2961 384

Page 32: 16.00 o6.2 d lillis

AWARDING SCHOLARSHIPS IN PHYSICS

Page 33: 16.00 o6.2 d lillis

RELATIONSHIP BETWEEN PERFORMANCE IN NCEA LEVEL 3 PHYSICS AND SCHOLARSHIP PHYSICS

Page 34: 16.00 o6.2 d lillis

DIFFERENTIAL ITEM FUNCTIONING (DIF)

Two or more groups of candidates, matched for ability, perform differently on a particular item.

Comparison Groups DIF

Male – Female Yes

European – Māori No

European – Pasifika No

European – Asian No

Page 35: 16.00 o6.2 d lillis

DIF: ILLUSTRATION(Item 1 from the 2010 Geography 90704 examination)

Page 36: 16.00 o6.2 d lillis

[email protected]

04 – 463 - 4251