Choose discriminating evidence without mistaking fit for truth

Define the evidence boundary

A useful virtual control plan distinguishes explanations by predicting what additional evidence would look like. A good fit to already inspected observations is not the same as successful prediction on an unused record. Calibration failure, recovery bias, population heterogeneity and genuine response changes can sometimes fit the same initial pattern.

Connect mechanism and measurement

The original candidate vectors are A=[1,1,1], B=[2,2,1], C=[2,2,3] and D=[3,1,2]. The first two coordinates are used for an unweighted squared-residual comparison; the third is withheld as a candidate prediction. These numbers are abstract interpreted responses, not conditions for a biological experiment or labels for particular agents.

For observations[2,2], B and C both fit exactly but predict different third values, one and three. That unused coordinate can distinguish these two supplied models if it is independently obtained with a suitable uncertainty bound. A third value two would instead challenge both exact models rather than requiring the analyst to force one to win.

A complete evidence plan defines the target quantity, source and permission, measurement specificity, independent units, missingness, uncertainty and competing predictions before feedback. This course permits supplied or existing documented records only, with no culture, chemical preparation, exposure or resistance generation. A best candidate among an incomplete set remains a bounded model judgment, not truth or practical competence.

Reconstruct the original record

At first observations[2,2], residuals are A2, B0, C0, D2. B and C tie, predicting unused coordinate 1 versus 3. Inspecting that coordinate first and then calling it an independent test would conceal feedback exposure and invalidate the intended separation.

Predict and investigate

Predict first: What makes an additional record discriminating rather than merely another fitted coordinate?

Calculate all candidate residuals, preserve ties and state the unused predictions before revealing a new record. Explain how uncertainty or a value outside every prediction would change the conclusion.

Defend a changed case

Apply the mechanism to a changed boundary

Use first observations[3,1]. Find residuals and the best supplied candidate. Does it prove that candidate is the true mechanism?

Retain the first response, units, uncertainty and any prior feedback exposure in the notebook before comparing this public explanation.

Compare a worked response

Residuals A4, B2, C2, D0 give D the unique best fit among the supplied set, with unused prediction 2. An omitted explanation, noise model or source error could still matter; exact fit is not universal truth.

Retain bounded evidence

Preserve the original source, first response, units, uncertainty and feedback exposure in the Microbiology evidence workspace. All numerical records are fictional. No culture, control-process execution, chemical preparation, genetic manipulation, resistance generation, participant testing or clinical decision is requested. Actual practical and independent learner competence remain separate.

Primary mechanism or evidence context, checked 31 October 2026. Original teaching and supplied records; no source operating procedures copied.