Jonas PowerPoint slides - CCSSO State Consortium on Educator

advertisement
Implementing Virginia’s Growth
Measure: A Practical Perspective
Deborah L. Jonas, Ph.D.
Executive Director, Research and Strategic Planning
Virginia Department of Education
May 2012
2
Potential Uses for Student Growth
Percentiles
• School improvement (including accountability) and
program evaluation
• One component of comprehensive performance
evaluation
▫ Consistent with the Code of Virginia requirement to
incorporate measures of student academic progress in
evaluations (§ 22.1-295).
▫ Growth percentiles may form the basis of one of multiple
measures of student progress, when available and
appropriate.
• Planning for differentiated teacher professional
development
• Communications with students and parents
3
Current Policy and Program Initiatives
• Implementing 2011 Virginia Board of Education
Guidelines for Uniform Performance Standards and
Evaluation Criteria for teachers statewide, which:
▫ includes a comprehensive model for teacher performance
evaluation;
▫ recommends that evaluations include multiple measures of
student academic progress; and
▫ recommends that one measure of student academic
progress be the state provided student growth percentiles
when available and appropriate.
• Incorporating growth measures into Elementary and
Secondary Education Act waiver application.
• Student growth percentile training for administrators.
4
Overview of our work
• Began working with nationally recognized
experts in calculating student growth percentiles
in spring 2010.
• Discovered that a standard
development/modeling approach would not
work in Virginia.
• Developed policy solutions—some of which may
be informed by growth data; many can be
informed by additional research.
5
Today’s topics
• Varied test progressions in statistical models
• Imperfect model fit—ceiling effects
• Changing tests in 2011-2012
• Other
6
Test progressions
• Not all students progress through Virginia’s tests
using the same pathway, e.g.,
▫ Grade 6, 7, Algebra I
▫ Grade 6, 7, 8, Algebra I
• This impacts statistical model building and the
“peer” or comparison group.
• States using or moving to end-of-course tests may
need to consider how to model unique student
progressions through courses and tests—they do
have an impact.
7
Model fit—Impact of floor and ceiling
effects
• Compromises validity of some statistical
estimates.*
• Important issues:
▫ Impact on individual student data.
▫ Impact on grouped data (teachers, grade-level
schools, districts).
• Policy should be developed to account for these
issues.
*See Koedel , C. & Betts , J. (2009) . Value-added to What? How a ceiling in the testing instrument influences valueadded estimation. http://economics.missouri.edu/working-papers/2008/WP0807_koedel.pdf.
8
Concrete example of ceiling effect
impact*
Scaled
Score
Grade 3
Scaled
Score
Grade 4
Scaled
Score
Grade 5
SGP
Grade 5
Student A
600
600
600
67
Student B
554
600
600
80
Student C
522
600
600
88
Scaled Score Range 200 - 600
9
Concrete example of ceiling effect
impact
Scaled
Score
Grade 3
Scaled
Score
Grade 4
Scaled
Score
Grade 5
SGP
Grade 5
Student A
600
600
600
67
Student B
554
600
600
80
Student C
522
600
600
88
This is the maximum
possible growth score
for students with a
perfect score for three
consecutive years.
Growth is still
considered high (in
Virginia), how high is
not clear.
10
Concrete example of ceiling effect
impact—a more extreme case
Scaled
Score
Grade 3
Scaled
Score
Grade 4
Scaled
Score
Grade 5
SGP
Grade 5
Student A
600
600
600
24
Student B
568
600
600
50
Student C
522
600
600
59
11
Concrete example of impact of ceiling
effects—a more extreme case
Scaled
Score
Grade 3
Scaled
Score
Grade 4
Scaled
Score
Grade 5
SGP
Grade 5
Student A
600
600
600
24
Student B
568
600
600
50
Student C
522
600
600
NOTE: The impact of significant floor effects would be similar.
59
This is the maximum
possible growth score for
students with a perfect
score for three consecutive
years, and may not be a
valid measure of student
growth.
This occurs because the
majority of students who
earn perfect scores for two
consecutive years also earn
a perfect score in the third
year.
12
What if you have floor or ceiling
effects in your tests?
• Statistical adjustments may be possible, depending on the
magnitude of the ceiling effects (Koedel & Betts, 2009).
▫ Statistical adjustments that may be acceptable for research
purposes are not always acceptable for accountability purposes.
▫ More statistical work is needed in this area.
• Policy for Growth Data
▫ Virginia established a cut-off for which students will receive and
not receive growth data.
▫ Resulted in the state providing no data for a substantial
percentage of students.
• Policies related to growth data cut-offs will be re-evaluated
when new tests are in place.
Koedel , C. & Betts , J. (2009) . Value-added to What? How a ceiling in the testing instrument influences value-added
estimation. http://economics.missouri.edu/working-papers/2008/WP0807_koedel.pdf.
13
Changing tests and growth data
• Virginia has not adopted the Common Core State
Standards
▫ Implemented new mathematics assessments this year, 2011-2012.
▫ New English assessments will be implemented in 2012-2013.
• New tests complicate growth model development,
training, interpretation, and timing of various use-cases.
14
Considerations when tests change
• Growth-to-Standard calculations
▫ It is not clear how many years of data are needed to measure growth-tostandard reliably.
▫ The amount of data needed depends at least in part on the magnitude of
changes in student outcomes each year.
• Use of growth data in school and educator accountability
▫ Virginia has not included growth-to-standard data in proposed school or
educator accountability due to changing tests.
▫ Statewide training is focused on data interpretation and use of multiple
measures to support teachers work to ensure students can meet new and
higher expectations.
▫ Regularly analyzing data and considering policy and programmatic
implications.
• Public information becomes an even greater challenge after tests
change.
Stay tuned……there’s more to learn!
15
A few other considerations
• Lack of growth data for students assessed with
alternative tests and the impact on use in evaluation and
accountability.
• Impact of floor/ceiling effects on other data derived from
state assessments (e.g., Lexile® and Quantile® data).
• Public information requests for data and Freedom of
Information Act/privacy laws and regulations.
• Data interpretations when growth is calculated for
assessments when more than one year is between last
and most recent tests.
Lexile® and Quantile® are trademarks of MetaMetrics, Inc., and are registered in the United States and abroad. Copyright
© 2012 MetaMetrics, Inc.
16
Contact Information
Deborah L. Jonas, Ph.D.
Executive Director for Research and Strategic
Planning
Virginia Department of Education
Deborah.Jonas@doe.virginia.gov
Download