Skip to content

Correct numerical calculations and quiz answer defects #9

Open

Description

@hananather

Question

Independent calculations found incorrect geometric support, a binomial mode outside support, wrong quiz numbers and duplicate correct options. The paired t-test example shows a two-sided alternative but uses a one-sided p-value, reversing its displayed decision.

Correct verified errors and necessary conditions while preserving question IDs. Add independent numerical reference fixtures, distribution invariants, answer-index validation, and a complete question inventory.

Acceptance: numerical fixtures pass, each single-answer question has one defensible key, assumptions are explicit, and the paired example decision follows its displayed test.

Parent: Probability Lab improvement roadmap.

Activity

  1. self-assigned this
    on Oct 3, 2026
  2. hananather commented on Oct 3, 2026

    @hananather
    OwnerAuthor

    Checkpoint 0433d85 in draft PR #19 adds independently checked central F-test cutoffs (two-sided alpha/2 tails, strict unrounded decisions), exact source-pair correlation summary calculations, and corrected quadratic/sine correlation teaching examples. SciPy/Fraction/analytical independent controls and actual-render cases passed within the clean 792-test gate. Empirical parameter changes now clear stale samples while preserving running/paused intent. Histogram density units and sample-size confidence wording are separate candidates under review. This does not certify unreviewed questions or content.

  3. hananather commented on Oct 4, 2026

    @hananather
    OwnerAuthor

    Checkpoint ca9a802 repairs sample confidence critical values, empirical density normalization, sample/population correlation notation, signed-square rendering, perfect-correlation limits, actual gallery labels and scoped model explanations. Clean suite: 955 passing tests plus one explicit provider-fixture skip; lint/build/110 page checks pass. Native scoped checks cover correlation's five main patterns, six galleries, 20 source pairs and keyboard-accessible formulas at phone and desktop widths.

    The sample-size budget page remains below the quality gate: its displayed optimum is grid-dependent (475 instead of the affordable integer 500 under its default model), infeasible settings can retain an old solution, E is labeled as currency, cost units are mislabeled, mount can award exploration, and enrollment/cost assumptions disagree. Independent probes reproduce seven cases; a scoped solver/units/retention/completion repair is in progress. This issue remains open.

  4. hananather commented on Oct 5, 2026

    @hananather
    OwnerAuthor

    Sample-size controls and explanations are repaired at cff80bc/8a850af (included in7bdb749). Empty, nonpositive and unrepresentable calculator inputs show a reason and disable derivation/save; keyboard buttons replace formula/preset click-only controls. Sample-size practice requires the exact integer, clears stale feedback on editing, preserves decimal margin-of-error answers, and explains the99% critical value correctly. Foundations hover uses unrounded critical values and rounds the required count up: native E=3→97 and confidence94.5%→208.

    Two passages now define E as the mean confidence interval's target half-width and show margin=z×standard error=zσ/√n. Native checks verify rendered fraction/root with no MathJax errors, calculator invalid/valid states, course preset139, integer82/noninteger81.5 rejection, clinical107 and decimal2.94 solutions. Keyboard access works at320/390 widths without document overflow. These are scoped repairs; mobile inner formula scrolling, canonical study/resume integration and the lesson's full teaching/visual gate remain open. No whole-curriculum rating or mastery claim is inferred.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions