SAT Scores (Highest Section Score)

Reporting year. For 2011-2024 this is the spring (ending) calendar year of the school year in the source's LONG_SCHOOL_YEAR; for 2004-2010 the source carries no year column and the filename publication year is used.

GOSA district code (FK to districts dimension): 3-digit zero-padded county/city codes or 7-digit state-charter codes. NULL on state-level rows.

GOSA school code, zero-padded to 4 characters (composite FK to schools dimension with district_code; not globally unique on its own). NULL on district- and state-level rows.

Student demographic (FK to demographics dimension). Breakdowns exist only for 2004-2010: race buckets (asian_pacific_islander, black, hispanic, native_american, white, other, race_unknown) and gender (female, male); every 2011-2024 row is 'all'. The source's bare 'Asian' label is the pre-1997 OMB COMBINED Asian/Pacific Islander bucket (no era of this source publishes a separate Pacific Islander row), so it maps to asian_pacific_islander — race buckets are mutually exclusive within this topic. 2007-2010 subgroup rows carry counts only (avg_score is NULL); race_unknown ('No Response') carries data in 2004 only (the source's R columns exist through 2007 but are blank in 2005-2006 and literal-'NULL' in 2007).

SAT section/composite, era-aware: 2004-2010 wide files publish reading (source 'High Verbal'), mathematics ('High Math'), verbal_math (their 400-1600 two-section total, equal to reading + mathematics) and writing (2008-2010); 2011-2016 old-SAT files publish reading, mathematics, writing and combined (the 600-2400 three-section total); 2016-2024 redesigned-SAT files publish combined_test_score (400-1600), math_section_score (200-800), reading_test_score / writlang_test_score (GOSA-rescaled 10-40 section tests, observed ~190-340), evidence_based_reading_and_writing (200-800, 2016-2019 only), and the essay components (essay_reading_score / essay_analysis_score / essay_writing_score on 2-8, essay_total on 6-24). Old-SAT and redesigned-SAT composites are NOT comparable across 2016.

Number of students tested. Float64 by documented exception (data-cleaning-standards §16): 2016-2024 bronze publishes FRACTIONAL counts on the combined_test_score component because GOSA reports it as the equal-weight mean of the three SAT section test-taker counts (Math, Reading, Writing & Language), not a headcount (e.g. (47+37+37)/3 -> 40.3 — exact thirds in 2016-2019, rounded to one decimal from ~2020); rounding to an integer would lose precision, and every other component's count is integral (enforced by a quality check). 2004-2010 sources publish one count per (entity, demographic) administration, repeated on each of that demographic's component rows; 2011-2024 sources publish true per-component counts. Either way, summing num_tested across test_component double-counts students — filter to one component for headcounts. A count of 0 is a real observation (zero students in that demographic took the SAT).

Average SAT score on the component's own scale (see test_component; scales range from essay 2-8 up to combined 600-2400, so never aggregate avg_score across components). Known source defects NULLed per data-cleaning-standards §4b (rows and num_tested preserved; 20 values): (a) 2009 Heritage High School (722:176) published reading=1022, mathematics=995, writing=976, verbal_math=2017 and South Atlanta Leadership (761:308) writing=811 — all above the 200-800 / 400-1600 scales; (b) 2010 Rockdale County High (722:3052) published reading=1055, mathematics=1113, writing=1009, verbal_math=2168 and Elberta Open Campus (676:3050) reading=822; (c) 2011-2015 published 10 old-SAT writing averages below the 200 section floor (7 school + 3 district rollup values, as low as 112). An average cannot leave the per-student scale, so these are publication errors; NULLing them revises the preserve-bronze default for this column and keeps the per-component range quality check enforceable. Preserved-but-flagged instead: 2010 Elberta verbal_math=1522 (in range, but its masked reading component makes it suspect) and the systematically low 2011-2016 writing averages (state 361-406 vs ~470 on the recent-SAT basis) — see limitations.