Bound an inferred quantity without hiding model error

1. Orient: Bound an inferred quantity without hiding model error

An inferred concentration depends on several inputs, not just the sample reading. In c=(y−b)/m, errors in signal, intercept and slope can all change the result. State whether each quoted uncertainty is a worst-case bound, a standard deviation, a confidence interval or something else; these meanings are not interchangeable.

2. Explain the mechanism and assumptions

Here the signal has bound±0.02AU, intercept0.1 has bound±0.01AU, and slope0.5 has bound±0.025AU/(mmol/L). For the positive numerator and slope in this model, the smallest possible concentration uses the lowest signal, highest intercept and highest slope. Reverse those choices for the largest. This monotonic argument produces an interval without inventing a probability distribution.

The displayed reference additionally removes a selected sample-only additive artifact. That artifact is deliberately absent from the declared error bounds. A narrow interval can therefore miss the reference even though its arithmetic is correct. Repeated scans might reduce some random component but would not necessarily identify a shared sample matrix effect or a wrong calibration model.

Report both the included uncertainty sources and the alternatives left outside the model. A useful next experiment changes an assumption in a controlled way: a matrix-matched reference, an independently measured sample quantity, or a comparison across a checked dilution range. Do not claim that interval overlap proves biological equality or that nonoverlap automatically identifies a mechanism.

3. Work a quantitative example

At signal0.6AU the nominal estimate is1mmol/L. The included-error lower bound is(0.58−0.11)/0.525≈0.89524; upper bound is(0.62−0.09)/0.475≈1.11579mmol/L. A hidden+0.15AU sample artifact makes the model reference(0.6−0.15−0.1)/0.5=0.7mmol/L, outside that interval.

4. Predict and investigate

Predict first: Can a correct calculation of the declared interval fail to contain the reference when the measurement model omits a bias?

Keep signal0.6AU and compare artifact0,0.1 and0.15AU. Identify which displayed bounds remain unchanged and why. Then change the signal while retaining the same artifact; explain the difference between absolute artifact and relative concentration error.

Retain the initial prediction, model settings, raw displayed quantities and your interpretation in the lesson notebook. All generated observations are explicitly synthetic; no wet-laboratory procedure is required.

5. Transfer to a changed case

Try a changed case before opening the response

For signal0.5±0.02AU, intercept0.1±0.01AU and slope0.5±0.025AU/(mmol/L), give worst-case lower and upper concentration bounds. Are these a95% confidence interval?

Write your calculation, units and model boundary in the retained lesson notebook first. This worked response is public practice, not independent assessment.

Compare with a worked response

Lower=(0.48−0.11)/0.525≈0.70476 and upper=(0.52−0.09)/0.475≈0.90526mmol/L. They are deterministic bounds under the stated input limits and model, not a95% confidence interval. Omitted sample interference can invalidate coverage.

6. Reflect and retain a defensible claim

Explain which result would weaken your favored mechanism. Distinguish the biological unit from repeated readings, included uncertainty from omitted bias, and a model calculation from a performed physical investigation. State one additional control or independent observation needed before generalizing beyond this case.

Source and course continuation

Scientific supporting reference, checked30October2026. All teaching wording, datasets, calculations and visualizations here are original; no source figure or exercise is reproduced. Return to General Biology I for the course-specific investigation, mixed written responses and staged mechanism project. This implemented category is the first of six planned modules, not the complete course or laboratory credit.