Evidence: DV13 and DV14
Module 7 — Dashboards, Ethics & Misleading Graphics
Make every metric answerable.
In this module, we will design dashboards that support informed decisions while clearly communicating denominators, differences between overall and subgroup patterns, uncertainty, missing or underrepresented groups, and privacy risks.
Work through one focused step at a time. The step navigator shows your place and lets you review earlier work.
Start Module 7 with the learning plan →
Short, focused steps
3 interactive audits
1 module graded assignment
50–65 minutes
Mission progress0%
Start by predicting where dashboards become statistically unsafe.
Dashboard reasoningKPI definitions, denominators, filters, aggregation, Simpson’s paradox, and decision rules
Ethical visualizationmisleading encodings, missing groups, sampling bias, minimum cell size, privacy, and inference boundaries
Evidence workflowInline browser R labs → GeoGebra prototype → AI audit → defense; optional desktop RStudio/Positron for Shiny
Evidence details: DV13 and DV14
Your evidence map
Evidence collected Demonstrated
DV13 and DV14 evidence details
Feedback
No evidence collected yet
Begin with the prediction checks. Practice changes the map to evidence collected; only final mastery marks skills demonstrated.
Next best action: prediction checks →Mission briefingView full topics and learning outcomes
What students will cover
- Metric definitions: numerator, denominator, reporting window, exclusions, eligibility rules, and the difference between counts and rates.
- Aggregation: filters, subgroup comparisons, Simpson’s paradox, unstable rankings, and when pooled summaries reverse the evidence.
- Dashboard design: decision questions, reactive controls, linked interpretation, uncertainty, accessibility, and cognitive load.
- Ethics and privacy: missing groups, sampling bias, small-cell disclosure, minimum cell sizes, color and encoding choices, and stakeholder impact.
- Inference boundaries: separating description, association, prediction, and causation in dashboard captions and AI-generated claims.
Student learning outcomes
- Define every dashboard KPI so another analyst can reproduce it.
- Detect aggregate patterns that reverse or conceal subgroup evidence.
- Choose controls that answer a real decision question rather than adding decorative interactivity.
- Enforce privacy protections and explain who may be missing or harmed by the display.
- Audit misleading graphics and revise unsupported AI claims.
- Build and defend a small Shiny, Quarto, or GeoGebra-enhanced dashboard prototype.
Module 7 · Step 1 of 6 · Watch8 min
Interactive reading: predict where dashboards become unsafe
Select True or False before opening the explanation. The goal is to notice statistical and ethical risk before polished design makes it feel trustworthy.
Denominator · predict before reveal
A dashboard shows “completion rate: 78%” but never defines the denominator. Is the tile interpretable?
No. Completion among enrolled, active, started, eligible, or assessed students are different metrics. A rate is not checkable until the numerator, denominator, window, and exclusions are visible.
Aggregation · predict before reveal
A district average can rise in the same year that every campus inside it falls, purely because the campuses changed size. True or false?
True. Changing subgroup sizes can raise an aggregate while every subgroup declines. Show the disaggregated pattern and explain the weighting before using the pooled result.
Privacy · predict before reveal
A filtered dashboard cell contains three students and shows no names. Publishing its 67% pass rate can still reveal how an individual did. True or false?
True. People familiar with the subgroup may infer individual outcomes. Minimum cell-size suppression protects privacy even when names are absent.
Interactivity · predict before reveal
Adding more filters always makes a dashboard more informative. True or false?
False. A control earns its place only when it changes a meaningful decision question. Decorative filters add cognitive load and may create unstable or disclosive cells.
Inference · predict before reveal
A dashboard association between tutoring and pass rate proves tutoring caused the difference. True or false?
False. The display supports an observed association unless the study design justifies a causal claim. Motivation, prior preparation, and selection into tutoring may confound the pattern.
Checking what this checkpoint still needs…
Module 7 · Step 2 of 6 · Guided10 min
Define the metric before displaying it
Audit six dashboard tiles for missing definitions, reversed aggregates, small-cell disclosure, or sound statistical design.
Open the DV13 dashboard-metric audit · about 12 minutes
The dashboard tile audit needs JavaScript.
Open the two-audience caption lab · about 8 minutes
Same number, two audiences
The final project asks you to communicate to non-technical readers without dropping the statistics. Practice the core move here: caption one verified tile twice — once for the superintendent, once for the methods appendix — keeping the denominator and the uncertainty in both.
The audience-caption lab needs JavaScript.
Checking what this checkpoint still needs…
Module 7 · Step 3 of 6 · Guided12 min
Diagnose the harm before choosing the repair
Distinguish arithmetic errors, misleading encodings, omitted people, privacy failures, and unsupported claims. Each failure needs a different repair.
Open the DV14 misleading-graphics gallery · about 12 minutes
Misleading-graphics gallery
The misleading-graphics gallery needs JavaScript.
Open the dashboard AI-output audit · about 10 minutes
AI-output audit
This AI audit needs JavaScript.
Checking what this checkpoint still needs…
Module 7 · Step 4 of 6 · Challenge12–15 min
Prototype one decision, then verify it
Explore denominator-driven ranking instability, prototype one meaningful threshold control in GeoGebra, and verify the metric definition in browser R.
Open the denominator-instability visual lab · about 10 minutes
Hands-on visual lab
Twelve campuses share one true pass rate. The league table still has a winner and a loser, and both are decided by the denominator.
This visual lab needs JavaScript.
Open the GeoGebra dashboard-prototype lab · about 15 minutes
GeoGebra lab
Use GeoGebra sliders as a dashboard-prototyping layer: one control, one decision rule, one verified metric.
Interactive Math with AI / GeoGebra strand
Embedded GeoGebra applet loading…
Open the runnable dashboard-verification R labs · about 15 minutes
R lab
R / Quarto / Shiny workflow
Build a Shiny mini-dashboard with reactive filters, denominator notes, suppression rules, and linked interpretation.
Browser R note: the inline verification lab runs base R plus ggplot2 (installed automatically on first run — one short wait, cached after that). Shiny is a desktop tool for RStudio/Positron — use it for the graded dashboard artifact if you choose that pathway; the required verification steps all run right here in the browser.
R Lab A — DV13: Pooled reversal
Your task The editor opens with the detail data frame already loaded: a separate synthetic fixture using the same three campus names, split by course level, where Riverbend wins both levels and still loses once pooled. Press Run, read the level-by-level and pooled tables, then defend which ordering your dashboard reports.
Lab complete when you can state which campus wins both course levels, what happens once the levels are pooled, and which ordering your dashboard reports and why. Pooled rates weight each course level by its tested count; subgroup rates condition on course level. These answer different questions. Choose the comparison from the decision question, not the preferred ranking.
R / Quarto / Shiny workflow
Press Run: the lab prints the level-by-level rates, the pooled rates, and the three reversal checks. It is a fixed synthetic teaching dataset with invented campus counts, not collected observations.
R Lab B — DV14: Suppression rule
Your task The editor opens with the same synthetic detail frame and a MIN_CELL threshold already loaded. Press Run and count what the rule hides, then state the rule a dashboard must publish. Five is an illustrative policy threshold, not a universal standard or privacy guarantee. Check related totals and filters for reconstruction; suppression does not delete observations from the analysis.
Lab complete when you can state the minimum-cell-size rule, how many cells it suppresses here, and what a reader sees in place of a suppressed cell.
R / Quarto / Shiny workflow
Press Run: the lab flags each cell SHOW or SUPPRESS and prints the suppressed count. Nothing needs copying.
Checking what this checkpoint still needs…
Module 7 · Step 5 of 6 · Independent18–22 min
Module 7 studio: complete the integrated dashboard-and-ethics defense
This is one required portfolio artifact that combines dashboard design with ethical audit and repair. Your work autosaves on this device, and the AI grader evaluates statistical validity, ethical safeguards, and decision value—not decorative polish.
How does grading work?
The AI score is formative feedback, not your grade. Use it to find gaps, revise, and resubmit — you have unlimited attempts, and only your best work matters. A low AI score simply means the rubric wants more detail or evidence somewhere; it is never locked in.
Your official grade comes from Dr. Roberts's review of your Canvas packet. The AI grader exists to coach you toward a complete, defensible packet before you submit it — read its feedback, strengthen the weak spots, and resubmit as many times as you like.
Where do I get my data?
The supplied campus counts are synthetic fixtures defined in the lab scripts; external retrieval is not applicable. Preserve the script version with your evidence.
For defining and checking a rate: the R lab editor already holds dashboard, three campuses with passed and tested counts. Run it to define a numerator, a denominator and a pass rate you can defend.
For the reversal and the suppression rule: the rate-definition frame above has one grouping variable and no cell under 50, so it cannot show a pooled-versus-subgroup reversal. Two dedicated labs already hold that data: R Lab A — DV13 opens with the fixed teaching frame detail preloaded — a separate fixture using the same campus names, split by course level, where one campus wins both levels and still loses once pooled — and R Lab B — DV14 opens with the same frame plus the minimum-cell-size rule that hides two cells. Press Run; nothing needs copying. It is a fixed synthetic teaching dataset with invented campus counts, not collected observations.
Rather not write R? The ranking-instability lab and the dashboard tile audit both report numbers you can quote directly in your defence.
Prefer your own dataset? You can use any built-in R dataset (run data() to list them) or add your own CSV under Your data in the R lab (files up to 500 MB stay in your browser). But you don't need to — the R lab dataset is ready to go.
Open the DV13 and DV14 Project 4 evidence checklists
Project 4 evidence items: DV13 and DV14
Open the submission decision check
The Module 7 submission decision check needs JavaScript.
Open the Module 7 AI-feedback workspace
Submit the Module 7 evidence artifact for formative feedback
Module 7 workspace loading…
Capstone writing autosaves as you work.
Checking what this checkpoint still needs…
Module 7 · Step 6 of 6 · Prove mastery8 min
Prove mastery and close the mission
Reflect on the dashboard evidence you can now defend, then earn a perfect mastery check to save DV13–DV14 completion.
Adaptive reflection
What I know / where I go next
Mastery check loading…
Module 7 record
Your saved Module 7 record needs JavaScript. You can also see it on your skills dashboard report card.
Project 4 connection DV13 and DV14 are Module 7’s evidence items. Project 4 combines them with DV11 and DV12 from Module 6. Open the Project 4 page to check saved evidence and download eligibility; saving mastery alone does not complete that evidence.