Middle Grades Assessment
Performance by Configuration
Marie Barry
Bari Erlichson
Peter Noehrenberg
Using a statistical model helps answer the question
whether configuration affects middle school
performance on state achievement tests
• Configuration type falls into three broad camps:
– “K-8 ES” (Kindergarten through 8th grade ES)
– “6-8 MS” (6th through 8th grade MS)
– “Other” (e.g. K-6 ES, 7-9 Jr. HS, etc.)
6/8/2011
Middle Grades Assess Perf by Config
2
K-8 ES Predominate in lower DFGs, while 6-8 MS
are more common in wealthier communities
35,000
Number of Students Grades 6-8
30,000
25,000
20,000
15,000
10,000
5,000
0
A
6-8 MS
B
CD
Other
DE
FG
GH
Configuration
Type
K-8 ES
I
J
District Factor Group
6/8/2011
Middle Grades Assess Perf by Config
3
To maintain a balance in the number of observations by
configuration, analysis was limited to DFGs A, B, and CD
Count of Students by District Factor Group by Configuration Type
Configuration Type
DFG 2000
A
B
CD
Config. Tot
6/8/2011
K-8 ES
Other
6-8 MS
DFG Total
115,429
(45.74%)
46,990
(29.18%)
23,945
(15.88%)
186,364
(33.03%)
69,332
(27.47%)
49,975
(31.03%)
71,459
(47.38%)
190,766
(33.81%)
67,610
(26.79%)
64,091
(39.79%)
55,409
(36.74%)
187,110
(33.16%)
252,371
(100.00%)
161,056
(100.00%)
150,813
(100.00%)
564,240
(100.00%)
Middle Grades Assess Perf by Config
4
Accounting for these independent factors
which are known to affect assessment
outcomes allows appropriate comparisons
• Poverty status (free or reduced-price lunch participation)
• Instructional Group (General Education, Special
Education only, Limited English Proficient only, but
excluding Special Education/Limited English Proficient)
6/8/2011
Middle Grades Assess Perf by Config
5
Data source and outcome measures
• Data relies on State assessments in grades 6-8 over
five years (2006-2010 inclusive)
• Language arts and mathematics sections were
modeled separately
• Outcome measure is the probability of achieving
proficiency/advanced proficiency on the ASK6
through ASK8
• “Treatment” variable (i.e. variable of interest) is
configuration type
6/8/2011
Middle Grades Assess Perf by Config
6
When holding lunch status constant there are no
significant differences* between configurations in
middle school language arts predicted pass rates**
Adjusted Overall Middle School Language Arts Percent
Proficient/Advanced Proficient by Lunch Status by Configuration
Percent Proficienct/Advanced Proficient
(with 95% confidence interval)
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
6-8 MS
Other
K-8 ES
6-8 MS
Other
K-8 ES
Lunch Participant
Lunch Participant
Lunch Participant
No Lunch
No Lunch
No Lunch
Upper CL
59.1%
60.4%
65.1%
78.4%
79.0%
77.6%
Lower CL
48.9%
50.9%
53.6%
69.6%
71.7%
67.1%
Pass Rate
54.0%
55.7%
59.5%
74.3%
75.5%
72.7%
*Statistical significance for the difference in the predicted pass rates is determined by whether their respective confidence interval
contains the other’s predicted pass rate; if each predicted pass rate lay outside the others’ respective confidence interval then the
difference is considered statistically significant at the given confidence level (95 percent in this case)
**“Predicted Pass Rate”: proficiency/advanced proficiency percentage after adjustment for other factors across grades and years
6/8/2011
Middle Grades Assess Perf by Config
7
When holding lunch status constant there are no
significant differences* between configurations in
middle school mathematics predicted pass rates**
Adjusted Overall Middle School Mathematics Percent
Proficient/Advanced Proficient by Lunch Status by Configuration
Percent Proficienct/Advanced Proficient
(with 95% confidence interval)
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
6-8 MS
Other
K-8 ES
6-8 MS
Other
K-8 ES
Lunch Participant
Lunch Participant
Lunch Participant
No Lunch
No Lunch
No Lunch
Upper CL
54.6%
55.5%
61.3%
73.1%
72.3%
71.6%
Lower CL
41.4%
44.5%
47.5%
60.5%
62.7%
58.5%
Pass Rate
48.0%
50.0%
54.5%
67.1%
67.7%
65.3%
*Statistical significance for the difference in the predicted pass rates is determined by whether their respective confidence interval
contains the other’s predicted pass rate; if each predicted pass rate lay outside the others’ respective confidence interval then the
difference is considered statistically significant at the given confidence level (95 percent in this case)
**“Predicted Pass Rate”: proficiency/advanced proficiency percentage after adjustment for other factors across grades and years
6/8/2011
Middle Grades Assess Perf by Config
8
Conclusions
• After controlling for numerous factors known to affect
proficiency on state assessments, school configuration
type was, by itself, not a significant factor
• School configuration has significant association with
lunch participation, but within each level of lunch status
there are no statistically significant differences across
configurations in language arts or mathematics
State Board Presentation
Middle Grades Assess Perf by Config
9