Sunteți pe pagina 1din 37

Advancements in Large-scale

Assessment
Heiko Rlke & Krisztina Toth
DIPF - Deutsches Institut fr Internationale
Pdagogische Forschung
Frankfurt am Main
TAO Days | 10.09.2012

Seite 1

Overview

Trends in LSA
Closer Look at Trends
Example: Making Use of Data
Challenges in LSA

TAO Days| 10.09.2012

Seite 2

Trends
Main trend: CBA
Sub-trends/effects of CBA:
CAT
towards formative assessment

Complex items
Simulation, interconnection

Interweaving
different tests, questionnaires, etc.

Big data
TAO Days | 10.09.2012

Seite 3

Trend: CBA
Computer-based Assessment has to serve a purpose
Costs?
Time!
Validity!

TAO Days | 10.09.2012

Seite 4

CAT
Also not practiced for its own sake
Not as important in pure summative assessment
But:
Strict time limits (e.g. PIAAC)
Growing demand for formative aspects in summative
assessment

TAO Days | 10.09.2012

Seite 5

Complex Items
Closer at reality
Simulation
Most important:
Ability to assess new domains

TAO Days | 10.09.2012

Seite 6

Examples of Complex Items

Simulation of an Email / Web Scenario


The test person receives an email and should book
cinema tickets online.

Dynamic Model (MicroDYN)


The test person should explore and master a dynamic
system with input (exogenous) variables influencing
output (endogenous) variables.

Automaton with Finite State Machine (MicroFIN)


The test person should interact with a mobile phone and set time to
summertime.

Interweaving of Tests and


Questionnaires
Framing
test only if preconditions are fulfilled

Double-check
findings of questionnaire

TAO Days | 10.09.2012

Seite 8

Big Data
Log data, not only results
Find out what is going on
Make use of data
E.g. partial scoring

-> Examples from


PIAAC PS-TRE

TAO Days | 10.09.2012

Seite 9

Simulation-based assessment
Modern assessments:
real life situations: e.g. web environments
complexity of instrument various ways
Monitoring students actions - log files
Analysis of test-taking paths is a relatively new field in education
(reason: printed vs. computer-based data collection)
Process data requires to consider and evaluate methods

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
10

Aim
to investigate how log data can be integrated into the process of
educational assessment and evaluation to support researchers
and practitioners in making use of the data assembled in
computer-based test delivery

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
11

Aim
to investigate how log data can be integrated into the process of
educational assessment and evaluation to support researchers
and practitioners in making use of the data assembled in
computer-based test delivery

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
12

Hypertext Item

13

14

Example log
829709 15
{"sender":"15","type":"loading","data":""}
833463 15
{"sender":"15","type":"loaded","data":""}
848153 15
{"sender":"15","type":"user_interaction","":"<?xml version=\"1.0\">
\u000d\u000a<cbaloggingmodel:EmbeddedLinkLogEntry
xmlns:cbaloggingmodel=\"http://www.softcon.de/cba/cbaloggingmodel\" id=\b576759f1:-7fed\" sourcePageId=\"Item15_linklist\"
targetPageId=\"Item15_website1\"
textFieldId=\"cbaTextField_71"/>\u000d\u000a"}
849094 15
{"sender":"15","type":"variable_change","data":{"name":
"snapshot_url,"value":"http://localhost:8101/cba-runtime/itemjsessionid=
LB1?custom_servicehandler=downloadService&file=
C:\\...snapshot5537437125227002523.xml"}}
855852 15
{"sender":"15","type":"user_interaction","data":"<?xmlversion=
\"1.0\encoding=\"UTF8\"?>\u000d\u000a
<cbaloggingmodel:ButtonLogEntryxmlns:cbaloggingmodel=\"http://www.soft
con.de/cba/cbaloggingmodel\id=\"cbaBackButton_14_13454937565947\"/>
\u000d\u000a}":"15","type":"unloaded","data":""}
15

Promising methods
Statistics and visualisation
Clustering
Classification

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
16

Process Measures
1.
2.
3.
4.
5.
6.
7.

Number of page visits


Number of different page visits
Visit of relevant page
Time spent on the relevant page
Ratio of time spent on the relevant page
Ratio of time spent on the opening screen
Completion time

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
17

Statistics
Process measures

Mean

Number of page visits


Number of different page visits
Time spent on the relevant page
Ratio of time spent on the relevant page
Ratio of time spent on the opening screen
Completion time

5.18
2.38
10.87
.14
.64
67.74

Standard
Deviation
3.13
1.38
9.77
.11
.18
27.49

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
18

Distribution of different page


visits

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
19

Distribution of different page


visits

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
20

Visualisation

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in


21
Large-Scale Assessments

Visualisation

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in


22
Large-Scale Assessments

Cluster Analysis
Features

Cluster 1

Cluster 2

Cluster 3

Cluster 4

23%

16%

34%

27%

Nr. of page visits


Nr. of different page visits
Completion time
Relevant page visited (Y/N)
Ratio of time on relevant page
Ratio of time on starting page
Distribution of sequences (%)

Ratio of correct responses

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
23

Cluster Analysis
Features

Cluster 1

Cluster 2

Cluster 3

Cluster 4

23%

16%

34%

27%

25.0%

1.7%

92.6%

84.4%

Nr. of page visits


Nr. of different page visits
Completion time
Relevant page visited (Y/N)
Ratio of time on relevant page
Ratio of time on starting page
Distribution of sequences (%)

Ratio of correct responses

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
24

Cluster Analysis
Features

Cluster 1

Cluster 2

Nr. of page visits

.72

5.41

Nr. of different page visits

.36

2.54

28.38

62.08

No

No

Ratio of time on relevant page

Ratio of time on starting page

.93

.60

Distribution of sequences (%)

23%

16%

25.0%

1.7%

Completion time
Relevant page visited (Y/N)

Ratio of correct responses

Cluster 3

Cluster 4

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
25

Cluster Analysis
Features

Cluster 1

Cluster 3

Cluster 4

Nr. of page visits

4.73

9.57

Nr. of different page visits

2.29

4.53

Completion time

59.60

91.60

Relevant page visited (Y/N)

Yes

Yes

Ratio of time on relevant page

.20

.17

Ratio of time on starting page

.62

.45

Distribution of sequences (%)

34%

27%

92.6%

84.4%

Ratio of correct responses

Cluster 2

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
26

Cluster Analysis
Features

Cluster 1

Cluster 2

Cluster 3

Nr. of page visits

5.41

4.73

Nr. of different page visits

2.54

2.29

Completion time

62.08

59.60

No

Yes

Ratio of time on relevant page

.20

Ratio of time on starting page

.60

.62

Distribution of sequences (%)

16%

34%

Ratio of correct responses

1.7%

92.6%

Relevant page visited (Y/N)

Cluster 4

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
27

Cluster Analysis
Features

Cluster 1

Cluster 2

Cluster 3

Cluster 4

Nr. of page visits

.72

5.41

4.73

9.57

Nr. of different page visits

.36

2.54

2.29

4.53

28.38

62.08

59.60

91.60

No

No

Yes

Yes

Ratio of time on relevant page

.20

.17

Ratio of time on starting page

.93

.60

.62

.45

Distribution of sequences (%)

23%

16%

34%

27%

25.0%

1.7%

92.6%

84.4%

Completion time
Relevant page visited (Y/N)

Ratio of correct responses

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
28

Summary and Conclusions

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
29

Summary and Conclusions


Features

Cluster 1

Cluster 2

Cluster 3

Cluster 4

Nr. of page visits

.72

5.41

4.73

9.57

Nr. of different page visits

.36

2.54

2.29

4.53

28.38

62.08

59.60

91.60

No

No

Yes

Yes

Ratio of time on relevant page

.20

.17

Ratio of time on starting page

.93

.60

.62

.45

Distribution of sequences (%)

23%

16%

34%

27%

25.0%

1.7%

92.6%

84.4%

Completion time
Relevant page visited (Y/N)

Ratio of correct responses

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
30

Future work
Pilot study sample size
Validation
Other types of items require new process measures

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
31

32

Future work
Pilot study sample size
Validation
Other types of items require new process measures
We have a lot of work to do:
Software developer
Test developer
Psychometricans

Berlin, 10.9.2012 | H. Rlke & K. Tth | TAO Days 2012 | Advancements in Large-Scale Assessments
33

(Selected) Challenges
Authoring and Management
Delivery
Re-use and exchange

TAO Days | 10.09.2012

Seite 34

(Selected) Challenges
Authoring and Management
Delivery
Re-use and exchange

TAO Days | 10.09.2012

Seite 35

(Selected) Challenges
Authoring and Management
Delivery
Re-use and exchange

TAO Days | 10.09.2012

Seite 36

(Selected) Challenges
Authoring and Management
Delivery
Re-use and exchange

TAO Days | 10.09.2012

Seite 37

S-ar putea să vă placă și