Note: The online Request to Speak window has expired.
Agenda Item
C.-1 24-2625 Presentation by the Superintendent of Schools, or designee to, and discussion with, the Board of Education, on Restructuring the District's Footprint including Optimal Location Analysis.
I can't speak to the soundness of the analysis, but I believe the major assumption underlying it is false. The presentation states, "the model facility capacity numbers used are the California Department of Education (CDE) recommendations for school size at each grade span." The CDE does not recommend optimal school sizes per se; although they reference numbers of pupils [500 (elem), 1,000 (middle), 2,000 (high)] in relation to minimum # acres needed per school site--10 (elem), 25 (middle), and 35-40 (high). This CDE "rule of thumb" adds one acre per 100 students above the minimum #. Also, according to CDE, "size of a school has been the subject of much research; there are no clear-cut solutions...the optimal school size is one that supports the kind of education the community wants at a cost it can afford; and the relationship between teachers and students is a primary concern. The school's structure, no matter what its size, must support that relationship."
The locations of students are generated from some probability distribution (not clear what distribution). The location optimization must therefore be repeated many times, using newly randomized points in every repetition, to prevent overfitting to the random points. The result will be a heat map showing better and worse locations for sites, rather than point locations that are purported to be optimal. This is Monte Carlo simulation. Further, the repetitions will show how sensitive is the model to the unknown locations of students.
Also, the analysis should show results when the constraints are relaxed (ie maximum number of schools, maximum number of students per school), in order to gain insight into the tradeoffs among the constraints.
A single repetition using a single set of constraints might just as well be cherrypicked.
I can't speak to the soundness of the analysis, but I believe the major assumption underlying it is false. The presentation states, "the model facility capacity numbers used are the California Department of Education (CDE) recommendations for school size at each grade span." The CDE does not recommend optimal school sizes per se; although they reference numbers of pupils [500 (elem), 1,000 (middle), 2,000 (high)] in relation to minimum # acres needed per school site--10 (elem), 25 (middle), and 35-40 (high). This CDE "rule of thumb" adds one acre per 100 students above the minimum #. Also, according to CDE, "size of a school has been the subject of much research; there are no clear-cut solutions...the optimal school size is one that supports the kind of education the community wants at a cost it can afford; and the relationship between teachers and students is a primary concern. The school's structure, no matter what its size, must support that relationship."
This analysis is flawed.
The locations of students are generated from some probability distribution (not clear what distribution). The location optimization must therefore be repeated many times, using newly randomized points in every repetition, to prevent overfitting to the random points. The result will be a heat map showing better and worse locations for sites, rather than point locations that are purported to be optimal. This is Monte Carlo simulation. Further, the repetitions will show how sensitive is the model to the unknown locations of students.
Also, the analysis should show results when the constraints are relaxed (ie maximum number of schools, maximum number of students per school), in order to gain insight into the tradeoffs among the constraints.
A single repetition using a single set of constraints might just as well be cherrypicked.