← All proof trees

PhiladelphiaAlgorithm

Citywide and peer-group percentiles

For each metric, year, and subgroup, every school is ranked among all default-included Philadelphia schools (citywide) and among its own peer group (peers). The percentile is the school's rank position, from 0 to 1, with ties sharing the midpoint rank. For metrics where lower is better (suspensions, chronic absenteeism), the ranking is flipped so that a high percentile always means a better outcome.

  1. 1
    How it was computedSteps
    1. Sort the comparison set by value; assign 1-based ranks with ties at the midrank.
    2. percentile = (rank − 1) / (n − 1); a set of one school gets 0.5.
    3. For lower-is-better metrics, invert (1 − percentile).
    4. Peer percentiles require at least 5 peers with a value; otherwise null.

    Inputs: every served cell; each school's peer group

    Evidence: compute_percentiles.ts — search "function rank"

  2. 2
    Assumptions & caveatsAssumptions
    • Suppressed or missing cells are excluded from the ranking (the school simply has no percentile that year).
    • Schools flagged include_in_default_comparisons = false are excluded from the citywide set.