F P Caillet Elementary · CDC 057905120 · Dallas ISD

Every published assessment record for one campus, 2012–2026

F P Caillet Elementary is a dual-language Dallas ISD campus where roughly half of Reading tests are taken on the Spanish form. This is the complete public STAAR and TELPAS record — captured directly from the Texas Assessment Research Portal, byte-validated, and analysed. Every number on this site is generated from the captured data, not transcribed.

What survives scrutiny

A detector suite produced 14,831 raw anomalies. Most were artefacts of how the data is shaped rather than anything about the school. Each correction below was measured, not estimated.

Stage	Count	Defect removed

The single most reliable conclusion is about the data, not the school: this campus is too small for most of the questions being asked of it. Typical cells hold 30–80 students, where nothing smaller than a 22–36 percentage-point change is detectable between two administrations. 76.5% of pre-declared equity contrasts show no detectable gap, and 62% of gaps larger than 30pp are small-sample artefacts.

Findings that hold

Finding	Evidence

What cannot be known from public data

Limit	Detail

Data explorer

Every performance-level row in the capture — … rows across programs, reports, administrations, subjects, grades, test forms, languages and demographic vectors. Filter and sort anything; the row count updates live.

Reading the bands. TEA publishes levels cumulatively: level 2 is “Approaches and above” and includes Meets and Masters. The n columns here are the exclusive bands, already derived, so they sum to the total. Populations below 5 students are suppressed by TEA and simply absent.

Anomaly findings

All … distinct findings, each tested against an explicit null model. Use the significance filter to see only what survives — the default view is everything, including what does not.

Why so few survive. Year-to-year variation here is about 2.5× what binomial sampling predicts, because each year is a different set of children. Retest administrations (May/June, grade 5) are populations of students who already failed, so comparing one against a primary administration is not like-for-like. Both corrections are applied.

Charts

Every cell carries its own label, and a table view accompanies each chart, so nothing depends on colour alone.

The source pages

Four pages of hand-written calculations were the starting point for this work. Each row was transcribed, then matched against the official record to identify which slice of data it described — year, grade, subject, language and test form are not written on the pages.

The pages are accurate. Where a row can be matched to a campus-level slice, it reproduces the official figures — several to the decimal. The deltas below are dominated by two things that are not arithmetic errors: TEA revised four cells after the pages were written, and some rows describe a classroom rather than the campus, which TEA does not publish at any grain.

Row-by-row reconciliation

Full transcription

Reports

Twenty-two analysis reports. Each was produced independently and then attacked by an adversarial review whose findings are included — several of the conclusions elsewhere on this site were overturned by it.