JC2 H3 Chemistry Tutor | Preparing a Residual Review Before Accepting a Quantitative Model

Begin with a specific claim: a proposed relationship r = k[A] adequately represents the supplied initial-rate observations over the stated concentration range. This is narrower than claiming that the reaction mechanism has been proved.

Supply k = 0.00800 s⁻¹ for the first comparison. Treat this as a proposed parameter value, not a least-squares result secretly inferred from the same table. The tutor should say where a value comes from.

State that r is the initial rate in mol/dm³/s and [A] is the initial concentration in mol/dm³. Other relevant conditions are held fixed in the fictional data. The dimensional check gives k units of inverse time.

The student’s job is to calculate predictions, compare them with observations and evaluate the claim at an appropriate level. A proposed alternative model or mechanism can be discussed later if enough information is supplied.

Write the intended conclusion in conditional language in the key. A residual pattern can reveal a mismatch worth investigating, while the strength of that finding depends on measurement information and the assumptions behind the comparison.

Contents · Next chapter

Use residual e = observed rate minus predicted rate. A positive residual means the observation lies above the model prediction; a negative residual means it lies below. The unit is the same as the rate unit.

Some sources use a different sign convention, so define it explicitly in the task. The important point is that the student applies one convention consistently and interprets the sign accordingly.

A residual is not a percentage error unless it is divided by an explicitly chosen reference and multiplied by one hundred. Keep raw differences separate from relative measures so that a sign or unit does not disappear unnoticed.

Ask the student to interpret one example in words before completing the table. If an observation is 1.00 × 10⁻⁴ and its prediction is 8.00 × 10⁻⁵, the residual is +2.00 × 10⁻⁵ mol/dm³/s.

Keep the observation and prediction in separate columns. Replacing the original rate with a corrected or rounded model value would erase the very evidence the review is intended to inspect.

In the original dataset, the signed residuals sum to zero: two positive differences of 2.00 × 10⁻⁵ balance the negative difference of 4.00 × 10⁻⁵. The mean signed residual is therefore zero despite the concentration-related pattern. Averaging signs can hide a structured mismatch.

Contents · Previous chapter · Next chapter

Supply initial A concentrations of 0.0100, 0.0200, 0.0300 and 0.0400 mol/dm³. The corresponding fictional observed rates are 1.00 × 10⁻⁴, 1.80 × 10⁻⁴, 2.40 × 10⁻⁴ and 2.80 × 10⁻⁴ mol/dm³/s.

With k = 0.00800 s⁻¹, the predicted rates are 8.00 × 10⁻⁵, 1.60 × 10⁻⁴, 2.40 × 10⁻⁴ and 3.20 × 10⁻⁴ mol/dm³/s. Check each multiplication independently before asking students to evaluate the model.

Subtract prediction from observation to obtain residuals +2.00 × 10⁻⁵, +2.00 × 10⁻⁵, zero and −4.00 × 10⁻⁵ mol/dm³/s. Retain signs because the direction of the mismatch contains information.

The final point has the largest absolute residual in this proposed comparison. Its absolute size is 4.00 × 10⁻⁵ mol/dm³/s. Absolute size and signed direction answer different questions.

The table should use a consistent power-of-ten scale or clearly written scientific notation. A formatting error that shifts one exponent can make the student diagnose a dramatic pattern that the original data never contained.

The mean absolute residual for the original proposed model is 2.00 × 10⁻⁵ mol/dm³/s. It reports an average deviation size without cancellation, but discards the order of the points. Keep it beside the full residual table rather than replacing the table with a single score.

Contents · Previous chapter · Next chapter

The lower-concentration observations lie above the proposed line, the third lies on it and the highest-concentration observation lies below. That structure prompts a closer look at the linear relationship and the comparison conditions.

Do not immediately announce a particular saturation mechanism. The residuals could be affected by the proposed parameter, a different functional relationship, measurement behaviour or uncontrolled conditions. The table alone does not distinguish all those possibilities.

Nor should the tutor call the highest point an outlier merely because it has the largest deviation. A point can reveal a systematic model limitation rather than an isolated measurement fault. Its treatment needs a reason supported by evidence.

If a residual plot is used in class, label the horizontal variable as concentration and the vertical variable as signed rate residual, with units and a zero reference. The exact table remains available so students can verify the plotted positions.

A useful first conclusion describes the pattern and proposes a check. For example, the proposed line overpredicts the highest-concentration rate, so examine the model and measurement response over that range before accepting it.

Ask whether reordering the rows changes the numerical residuals. It does not, but reviewing them against concentration reveals structure that an arbitrary row order may hide. The horizontal variable in a residual plot should connect to the model question, rather than simply use question number by default.

Contents · Previous chapter · Next chapter

Initially no measurement uncertainty has been supplied. The student can describe the residual pattern, but cannot give a formal statistical rejection or a probability that the model is wrong.

For a separate illustrative scenario, state that each observed rate has a known absolute error bound of 5.00 × 10⁻⁶ mol/dm³/s, while concentrations and the proposed parameter are treated as exact for this comparison. This is a supplied bound, not an inferred confidence interval.

Several residuals then exceed that rate-error bound. Under those restrictive assumptions, the proposed predictions cannot all agree with the observations within the supplied error limits. The mismatch concerns the combined model and comparison assumptions.

If the bound were instead 5.00 × 10⁻⁵, all four residuals would fall within it. That makes the proposed predictions compatible with those much wider bounds, but does not prove the relationship or a mechanism.

The table below uses the original data without attaching an uncertainty interpretation. The tutor can add either clearly labelled scenario after the basic residual calculation is understood.

Input uncertainty can also affect a prediction. For this linear model with k treated as exact, an illustrative concentration error bound of 0.00100 mol/dm³ gives a prediction bound of 8.00 × 10⁻⁶ mol/dm³/s. The observed-rate bound alone would then be an incomplete comparison.

If that input bound is combined conservatively with the illustrative 5.00 × 10⁻⁶ observed-rate bound, the maximum discrepancy allowed by those separate bounds is 1.30 × 10⁻⁵ mol/dm³/s. This is a worst-case bound calculation under supplied assumptions, not a statistical confidence statement.

Contents · Previous chapter · Next chapter

A student may suggest changing k. That is a reasonable proposal to investigate, but the fitting rule must be stated. A value chosen by eye, a value estimated from one point and a least-squares value answer different parameter-estimation questions.

Do not call 0.00800 the best-fit slope of this table. It was supplied as the proposed value. If a fitted value is introduced, retain a record of the data and criterion used to obtain it.

After adjustment, calculate the residuals again rather than assuming that a new slope removes every pattern. Changing a parameter can improve agreement while leaving systematic structure that suggests the model needs further examination.

A model assessed on the same observations used to choose its parameters has not yet passed an independent predictive test. Suitable additional data can provide a more informative check, particularly where competing models make different predictions.

Keep this at a supplied-tool level when the student has not studied regression methods. The task can ask what should be checked next without requiring an unannounced statistical calculation or specialist software routine.

A high fit statistic supplied in another task should be treated as one summary of agreement under its definition. It does not by itself establish correct units, stable conditions or a unique mechanism. Ask what information the statistic retains and what the residual pattern can still reveal.

Contents · Previous chapter · Next chapter

If concern centres on the high-concentration response, propose suitable additional concentrations and repeats in that region while preserving the relevant conditions. Explain what pattern would support or weaken the proposed relationship.

If concern centres on an instrument’s conversion of signal to rate, an appropriate calibration check addresses a different issue. Repeating the same calculation cannot resolve a mistaken measurement relationship.

If a time sequence is used to obtain initial rates, check that the same method estimates the initial slope for all concentrations. Mixing early-time and later average rates can create a misleading model comparison.

Ask students to connect each proposed check to a hypothesis. More data is too vague unless the student identifies where, under what conditions and which ambiguity it would help resolve.

The paper exercise does not select an actual apparatus or authorise a practical procedure. Its purpose is to teach how an evaluation question can lead to a specific evidence-gathering decision.

Contents · Previous chapter · Next chapter

Supply a new proposed model r = 0.0200[A], with concentrations 0.0100, 0.0200 and 0.0300 mol/dm³. Give observed rates 1.90 × 10⁻⁴, 4.10 × 10⁻⁴ and 5.90 × 10⁻⁴ mol/dm³/s.

The predictions are 2.00 × 10⁻⁴, 4.00 × 10⁻⁴ and 6.00 × 10⁻⁴. Residuals are −1.00 × 10⁻⁵, +1.00 × 10⁻⁵ and −1.00 × 10⁻⁵ mol/dm³/s under the defined convention.

Ask whether these differences alone establish random error. They do not. Three small alternating residuals can be described, but a statistical claim needs appropriate assumptions, measurement information and sufficient evidence.

Now supply an illustrative rate-error bound of 2.00 × 10⁻⁵ with other quantities treated as exact. Each residual lies inside it. The appropriate statement is compatibility with that supplied bound over these points, rather than proof of universal validity.

Record whether the student can compute the sign and write the conditional conclusion independently. A correct subtraction followed by an unjustified mechanism claim reveals a different teaching need from a simple exponent error.

For a near-zero predicted rate, a relative residual can become very large or undefined even when the absolute discrepancy is small. If relative measures are introduced, define their denominator and handle such cases explicitly. The raw residual retains a clear rate unit and remains interpretable.

Contents · Previous chapter · Next chapter

If a student says the model is wrong because no point matches exactly, ask what measurement information has been supplied and whether exact agreement is the intended criterion. Quantitative evaluation needs a stated basis for acceptable agreement.

If the student says the graph looks close so the mechanism is proved, ask which mechanism alternatives and measurement assumptions have been tested. Agreement with a numerical relationship is a narrower conclusion.

If the student deletes the largest residual, ask for the independent evidence that the observation is faulty. Retain the value while investigating; removing it solely to improve agreement creates a misleading record.

In group teaching, ask each student for one numerical residual and one permitted conclusion before discussing the full review. This preserves evidence of individual interpretation while allowing useful comparison of proposed next checks.

A parent update can identify the developing skill: the student now distinguishes a calculated deviation from a statistical conclusion and can propose evidence that would address the remaining uncertainty.

Contents · Previous chapter · Next chapter

Keep raw observations, model definition, parameter source, checked predictions, residual convention and any uncertainty scenario together. Label each scenario so its assumptions are not carried accidentally into another question.

Verify units and numerical scales before printing or using a spreadsheet. If software is used, hand-check one row and inspect whether cell references hold the intended parameter fixed.

Choose repair for unit or sign confusion, stabilise through a fresh supplied model and extend through parameter adjustment or a designed comparison when the foundations are secure.

Keep sensitivity analysis distinct from residual review. Sensitivity changes inputs within a model to inspect output behaviour; residual review compares model predictions with observations. Both can inform an argument, but they answer different questions.

The completed preparation gives the student a measured way to assess a quantitative claim: calculate the difference, inspect its structure, state the comparison assumptions and identify the next useful evidence.

A concise student conclusion can name the model, range and limitation: over the supplied concentrations, the highest observation lies below the proposed prediction; the uncertainty assumptions determine how strongly that mismatch counts against the comparison. The next check should address a specific model or measurement question.

Contents · Previous chapter · Continue to the Chemistry library