Title: Cell Size and Reliability Strategies
1Cell Size and Reliability Strategies
- CCSSO Mega-SCASSJanuary 22-23, 2005
- Facilitator Andra Williams
- Presenter Ellen Forte Fast
2Overview
- Why worry about AYP reliability?
- What affects AYP reliability?
- How are states addressing reliability in AYP?
- Where can I get more information?
3Why worry about AYP reliability?
- To reduce the likelihood of misclassifying
schools and districts - Most states are more concerned about incorrectly
identifying schools for improvement than about
failing to identify schools that really do need
help - To reduce the number of schools and districts
identified for improvement
4What affects AYP reliability and what can be done
about it?
- As the reporting statistic is reduced (e.g.,
changed from the mean to percent proficient),
reliability decreases - What states can do nothing right now, since
states must use percent proficient - As the number of scores increases, total score
reliability increases - Increase the minimum n and/or account for at
least some portion of the error
5How are states addressing reliability in AYP?
- Minimum n the number of scores required for
the cell to be considered for AYP - Confidence Intervals
6How are states addressing reliability in AYP?
- Minimum n the number of scores required for
the cell to be considered for AYP - Single year minimum
- Possible combination of data from 2-3 years
- Conjunctive total and by grade
- Proportion of total
- Different minimums for different groups
- Different minimums for different indicators
7How are states addressing reliability in AYP?
- Single year minimum
- Status
- Range 5 to 60
- Mode 40
- Trend
- Many states increased the minimum n
- Oklahoma changed from 30 to 0 with a 99 CI
8How are states addressing reliability in AYP?
- Possible combination of data from 2-3 years
- Status
- Some states allow combination as an option if the
single year n is below the minimum - Some states allow combination when AYP is missed
- Trend
- Shift from combination as a requirement to
combination as an option - California deleted the combination option
9How are states addressing reliability in AYP?
- Conjunctive total and by grade
- Kentucky originally required10 students per
grade and 30 students per school,now requires10
students per grade and 60 students per school
10How are states addressing reliability in AYP?
- Proportion of total
- Status
- A few states require the group to represent a
specific proportion of the school - Texas 50/10/200
- California 50/15/100
- A number of states require the group to represent
a specific proportion of the district - California 100/15/200
- Trend
- Increasing use of proportions for
districtse.g., minimum 40 for 1-4000 scores,
then 1 of scores
11How are states addressing reliability in AYP?
- Different minimums for different groups
- Status
- Many states use a larger minimum n for students
with disabilities - Some states use a larger minimum n for English
language learners - Maryland counts a student in only one program
category - Trend
- Increasing use of this strategy
- Increasing the minimum n
12How are states addressing reliability in AYP?
- Different minimums for different indicators
- Status
- A number of states use higher minimum n for
participation rate than for percent proficient
(range 30-50 mode 40) - Nevada for the participation rate, minimum
20,but if a school has fewer than 20 scores, n-1 - Trends
- Not much has changed other than taking advantage
of the March 29, 2004 guidance regarding the use
of data for 1, 2, or 3 years, whichever provides
the best result
13How are states addressing reliability in AYP?
- Confidence Intervals
- Statistical estimates of the range in which a
school or districts true AYP score (e.g.,
percent proficient) might fall - Address the question Is the AYP score really
different from the AYP target? - Balance between
- Falsely identifying good schools as needing
improvement - Failing to identify needy schools
Increase the size of the confidence interval
Decrease the size of the confidence interval
14How are states addressing reliability in AYP?
- Percent Proficient Many states use confidence
intervals of 95 or 99 - Safe Harbor Originally not allowed now many
states allowed to use a 75 CI for safe harbor
15Where can I get more information?
- National Center for the Improvement of
Educational Assessment - Hill, R. and DePascale, C.A. (April 2003).
Adequate yearly progress under NCLB Reliability
considerations. http//www.nciea.org/publications/
NCMEconf_RHCD03.pdf. - Richard Hill. (2000). Issues related to the
reliability of school accountability scores. - Richard Hill Charles DePasquale. (2002).
Determining the reliability of school scores. - CRESST
- Linn, R. L., Baker, E. and Betebenner, D. (2002).
Accountability systems Implications of
requirements of the No Child Left Behind Act of
2001. CSE Technical Report 567.
http//www.aera.net/pubs/er/pdf/vol31_06/AERA31060
2.pdf. - CCSSO
- Marion, S., White, C., Carlson, D., Erpenbach, W.
J., Rabinowitz, S. and Sheinker, J. (2002).
Making valid and reliable decisions in
determining adequate yearly progress.
http//www.ccsso.org/content/pdfs/AYPpaper.pdf. - Forte Fast, E. Hebbler, S. (2004). A Framework
for Examining Validity in State Accountability
Systems. http//www.ccsso.org/publications/details
.cfm?PublicationID236