Reproduce an analysis without overwriting its source records

Define the decision and evidence boundary

A reproducible environmental analysis needs identifiable records, units, provenance and a declared transformation recipe. An edited spreadsheet without the original values cannot show what changed. Keeping raw data separate from corrected values allows another analyst to recover the result and challenge its assumptions.

Reconstruct the transparent method

The original fictional packet cenv-basin-pairs@1 contains eight sites in four independently randomized treatment/comparison pairs. Before and after values are representative interval means, not repeated independent sites. Each record has a stable ID. A separate calibration statement stipulates a +2 mg/L after-reading offset only for BASIN-P3-T; the corrected value is 30−2=28.

Interpret the comparison

For each pair, compute treatment change minus comparison change. Raw contrasts are−4,−2,0,−2 mg/L, averaging−2. Correcting the known sensor offset gives−4,−2,−2,−2, averaging−2.5. The correction is not an outcome-driven deletion; its reason and affected record remain visible in the downloadable CSV.

Predeclare limitations and revision

Converting mg/L to µg/L multiplies all concentrations and differences by 1000. Counts, ratios and the physical interpretation do not change. Real calibration uncertainty, missing values and access controls would need explicit management. This packet is synthetic and must never be presented as a field study completed by the learner.

Worked original packet

BASIN-P3-T has before 28, raw after 30 and corrected after 28 mg/L. Its comparison changes 26→28, so the corrected contrast is(28−28)−(28−26)=−2. Across four pairs the mean changes from−2 to−2.5 mg/L.

Investigation

Download and inspect the raw/corrected CSV. Recompute each contrast from record IDs, toggle only the documented correction, then convert units. Preserve the packet identity and an explicit analysis recipe in the notebook.

Changed-case transfer

Express the corrected mean contrast in µg/L. Does the conversion create more independent evidence?

Feedback

−2.5 mg/L becomes−2500 µg/L. There are still eight sites and four independent pairs; unit conversion creates neither new observations nor stronger causal evidence.

Sources and evidence boundary

Original prose, dataset and numerical exercises. Primary method/mechanism source, checked 30 October 2026. The versioned packet is synthetic, not observed fieldwork. No treatment implementation, site access, pollutant handling or real management authorization is requested. Preserve first reasoning and critique in the course workspace.