Student Learning Data & My Evaluation For Instructional Personnel Spring, 2012
Presenters & Questions • Boyd Karns • Submit questions today • Jason Wysong • Submit questions online • Brandon McKelvey • Call/email us! – Boyd 5-0198 – Jason 5-0212 • Talk with your administrator
Today’s Focus • Requirements of Senate Bill 736 • How Florida will measure student learning – Concept – Example • SCPS Business Rules for 2011-12 • Plan for 2012-13 and beyond
Disclaimers • SCPS did not create the value-added model • FL value-added is different from other places • More flexibility for 2011-12 than 2012-13 • Every district using different rules for 2011-12
Annual Evaluation Ratings Highly effective Effective Needs Improvement/Developing* Category I: first 3 years—developing Category II: 4+ years—needs improvement Unsatisfactory
Underlying Philosophy Teachers are the single most important variable in a student’s academic growth. Teachers who effectively implement research- based practices will create student learning growth.
State Process • Method for measuring student learning growth on FCAT must be established by Florida DOE, with measures on other assessments to follow • DOE Student Growth Implementation Committee – Recommend a formula for learning growth measurement – Teachers, administrators, and university professors were appointed to this group – The DOE also contracted with AIR (American Institutes for Research) to provide technical assistance
Key Points • Growth, not proficiency • Learning growth, not learning gain
State Committee Recommendations • The Student Growth Implementation Committee recommended a covariate adjustment model – Covariates are also called variables and represent student characteristics which influence learning – This model yields a VALUE-ADDED score • This model establishes a personal learning growth expectation for all students in the state • If a student meets or exceeds their growth expectation, that student will positively impact their teacher’s value added score
Variables A value added model measures the impact of a teacher on student • learning by accounting for other factors that may impact the learning process The Student Growth Implementation Committee chose to include in the • model the following student characteristics that may influence a student’s expected growth – Number of subject-relevant courses – Two prior years of achievement scores – Student with Disabilities (SWD) status – English Language Learner (ELL) status – Gifted Status – Student Attendance – Mobility (number of transitions) – Difference from Modal Age in Grade (retention) – Class Size (number of students) – Homogeneity of Student’s Prior FCAT Scores
Variables not in the model • Gender • Race/Ethnicity • SES
School Component In addition to student-level scores, the model also calculates a • ‘school component’ – The ‘school component’ is actually a ‘grade-level, subject’ component – For example, all 5 th grade reading teachers at a school will have the same ‘school component’ score The school component is combined with the teacher calculation in • the value-added model – The Student Growth Implementation Committee chose the school component because they believed that there were school-level and classroom level factors that influence student learning – The committee thought that teachers should not be held completely responsible for student performance because some responsibility is held by the school as a whole
School Component • Elementary: 4 school components • Middle: 6 school components • High: 2 school components, maybe 3 • School components can vary significantly by grade level, subject, and from year to year • No direct link to school grades • No way to game the system
Implications • The value-added model starts by comparing students to others around the state. • The teacher’s initial score is the average of these student-level comparisons across the state. • The teacher’s score is adjusted based on the average performance of other students in the same grade level at the school. • This model is designed to control for school effects (leadership, climate, etc.)
Finding a Value-Added Score There are two major components of the value-added score • – Teacher Score (how effective is the teacher) – School Score (how effective is the school) The difference between the school and teacher score is called the • ‘teacher effect’ – This is the difference between a teacher’s effectiveness and the effectiveness of other teachers in the same grade-level and subject In order to find the value-added score, half of the school score must • be added back to the score – This is because the student growth committee chose to only use half of the school component score
Simple Example • Teacher score: 20 • School score: 10 • Unique teacher effect: 10 • Add ½ of school score back in: 15
Standardizing & Aggregating Scores • Since the average FCAT growth rate is different at each grade level, scores must be STANDARDIZED so that teachers of different grade levels can be compared. – This is done by dividing each teacher’s score by the average amount of growth at a grade level – This ‘smoothes’ out differences in growth at grade levels • This accounts for grade-level differences in FCAT.
Standardizing & Aggregating Scores • Since most teachers have students in multiple grade levels, value-added scores must be AGGREGATED so that each teacher receives only one overall score. – This is done through weighted averaging, so that proportion of students in each grade level is incorporated into the overall score • Standardization & aggregation allow for comparison of all teachers in the model regardless of subject, grade level, etc.
Standard Errors • All statistical measures contain a degree of error • Value-added scores have a ‘standard error’ – The standard errors are calculated by the DOE in conjunction with AIR • The standard error is a value that represents the amount of uncertainty that we have in a particular value – For our purposes, a higher standard error would suggest that we have less confidence in the score
Why Does the Standard Error Matter? • If we did not use the standard error in placing teachers in categories, we would be ignoring an important piece of information about the data • No data are perfect, but we have methods for determining how likely data are close to the ‘real’ value – Using data without this adjustment is not appropriate – Example: No one would sample two people in a Presidential poll and not mention that the ‘Margin of Error’ would be 99%
Teacher Value Added Scores at School X in 7 th Grade The dots above the teacher labels are teacher value-added scores. The lines extending from the bars are confidence intervals at 0.5 Standard Errors (SE). At 0.5 SE , Teacher B is lower and Teacher C is higher than the state mean. Teacher C 0.5 SE = 38% Confidence Interval Teacher B Teacher A -30 -20 -10 0 10 20 30 State Mean 7
Teacher Value Added Scores at School X in 7 th Grade At 1 SE , Only Teacher C would be considered significantly higher or lower than the state mean. 1 SE = 68% Confidence Interval Teacher C Teacher B Teacher A -30 -20 -10 0 10 20 30 State Mean 8
Teacher Value Added Scores at School X in 7 th Grade At 2 SE , none of the three teachers would be significantly higher or lower than the state mean. Teacher C 2 SE = 95% Confidence Interval Teacher B Teacher A -30 -20 -10 0 10 20 30 State Mean 9
Standard Error Implications • When you account for the standard error, you are able to have more certainty concerning which evaluation category is most appropriate for a teacher – A 95% confidence interval is built by adjusting for approximately 2 standard errors • However, the more adjustment that is made for the standard error, the wider the range of possible scores are created for each teacher – This means that most teachers will fall around the mean in a single category
Standard Error & Policy Use of standard error makes it more difficult to clearly differentiate teacher performance level…but this is exactly what 736 requires.
VAM Procedures for 2011-12
SB 736 in 2011-12 • 736 requires State Board of Education to establish business rules and set cut points for personnel evaluations • State Board will not set rules until 2012-13 • DOE required districts to establish their own rules for 2011-12
SCPS Process • Teacher Evaluation Committee • Teacher focus groups • Administrator Evaluation Committee • Dr. Vogel & District Administrators
SCPS Decision # 1 Use only 2011-12 student data – This is year 1 – No historical data – Reduces learning growth from 50% to 40% – Remaining 60% is based on supervisor evaluation
SCPS Decision # 2 • Use 2 standard errors with all value-added scores • Adjusting for 2 standard errors greatly increases the range of scores that influence a teacher’s placement • This suggestion is supported by educational research and prior ‘best practices’ using value- added modeling
Recommend
More recommend
