Sensitivity and Parameter Change: Ask What Changes When the Number Changes
Parameters do more than fill formulas. They control shape, location, number of solutions, rates, extrema and transitions between different mathematical behaviours.
Secondary 3 Additional Mathematics often asks for one parameter value. A deeper skill is understanding how the whole system changes as that parameter varies. Increasing a quadratic coefficient changes width and opening. Moving a constant can create or remove real roots. Changing an exponential rate changes growth speed. Trigonometric coefficients change amplitude and period. A model parameter can move an optimum or alter whether a solution is feasible.
This guide develops sensitivity reasoning: identify which features depend on a parameter, find transition values where qualitative behaviour changes, and compare small parameter changes without recomputing everything blindly.
AI Extraction Box: The Sensitivity Loop
- Identify parameter: which symbol controls the family?
- Identify outputs/features: roots, vertex, gradient, period, growth, optimum.
- Vary one parameter: hold others fixed.
- Find thresholds: solve the equality case where behaviour changes.
- Compare regimes: below threshold, at threshold, above threshold.
- Interpret direction: what increases, decreases, appears, disappears or shifts?
- Check domain: some parameter values may invalidate the original model.
Quadratic Sensitivity: Leading Coefficient
For y=a(x−h)²+k:
- a>0 → opens upward;
- a<0 → opens downward;
- larger |a| → narrower/steeper graph;
- smaller |a|, non-zero → wider graph;
- h and k still locate the turning point.
The parameter a controls curvature direction and scale without moving the vertex.
Worked Example 1: Compare a Family
Compare y=(x−2)²+1, y=3(x−2)²+1 and y=−2(x−2)²+1.
- All have turning point (2,1).
- a=1 gives the base width.
- a=3 gives a narrower upward parabola.
- a=−2 reflects downward and narrows relative to a=−1.
One parameter changed; one family feature stayed invariant while others changed.
Quadratic Root Transitions Through the Discriminant
For x²+kx+4=0:
Δ=k²−16.
- |k|>4 → two distinct real roots;
- |k|=4 → one repeated root;
- |k|<4 → no real roots.
The values k=±4 are transition thresholds. At them, the graph changes from crossing the axis twice to touching it once to missing it entirely.
Threshold values are where a qualitative feature appears, disappears or changes type.
Sensitivity of a Vertex to Parameters
For y=a(x−h)²+k, h controls horizontal position and k controls vertical position. Changing h shifts the graph without changing shape; changing k raises or lowers it without changing shape.
This lets students predict the effect before plotting.
Exponential Sensitivity
For P=Ae^{kt}:
- A changes the initial scale P(0)=A;
- k>0 gives growth;
- k<0 gives decay;
- larger positive k gives faster growth;
- more negative k gives faster decay.
A small change in k can produce a large long-term output difference because the parameter sits in the exponent.
Worked Example 2: Growth-Rate Sensitivity
Compare P₁=100e^{0.02t} and P₂=100e^{0.05t}. Both start at 100, but P₂ grows faster. At t=10:
- P₁=100e^{0.2};
- P₂=100e^{0.5}.
The difference between rate parameters 0.02 and 0.05 compounds over time.
Trigonometric Parameter Sensitivity
For y=A sin(Bx)+D:
- |A| controls amplitude;
- sign of A reflects vertically;
- |B| controls period 2π/|B|;
- larger |B| means shorter period;
- D shifts the midline.
These parameter effects are predictable before any detailed sketch.
Worked Example 3: Period Change
Compare y=sin x, y=sin2x and y=sin(x/2).
- sin x has period 2π;
- sin2x has period π;
- sin(x/2) has period 4π.
The input coefficient changes horizontal frequency inversely.
Sensitivity of Tangency Conditions
A moving line y=mx+c can change from two intersections with a quadratic to one tangent intersection to no intersections as m or c changes. The threshold occurs where the intersection discriminant equals zero.
This is a general pattern:
parameter change → discriminant changes sign → number of real intersections changes.
Calculus Sensitivity: Moving Stationary Points
Consider the family f(x)=x²−2kx. Then:
f′(x)=2x−2k=0 → x=k.
The stationary x-coordinate moves exactly with k. Completing square confirms:
f(x)=(x−k)²−k².
So the minimum point is (k,−k²). One parameter change shifts both location and value in a predictable way.
Worked Example 4: Parameter Changes an Optimum
Let A(x)=kx−x² for k>0. Then:
A′=k−2x=0 → x=k/2.
Maximum value:
A(k/2)=k²/4.
Doubling k doubles the optimal x but multiplies the maximum value by four.
This is sensitivity reasoning: study how the output changes as the parameter changes.
Parameters Can Change Feasibility
A model may only be feasible for certain parameter values. Suppose a dimension is x=10−k. If physical length requires x>0, then k<10.
A parameter may therefore control not only size but whether the model makes sense at all.
Small Parameter Change, Large Effect
Not every relationship responds linearly. In exponentials, a small change in rate can compound. Near a discriminant threshold, a tiny parameter change can change the number of real roots. Near a physical boundary, a small parameter change can make a previously feasible solution impossible.
Students should therefore distinguish:
- small numerical change;
- small effect on output;
- qualitative regime change.
These are not the same.
Sensitivity Through Graph Families
Sketching several members of a parameter family on the same conceptual axes helps students see invariant and changing features.
- Which intercepts move?
- Which asymptotes move?
- Which turning points move?
- Does the graph change orientation?
- Does the number of roots change?
- Does the range change?
This builds stronger parameter intuition than solving isolated k-values only.
Transition Thresholds
Important threshold conditions include:
- Δ=0: root-count transition for quadratics;
- parameter=0: growth/decay direction may change;
- amplitude=0: sinusoidal variation collapses to the midline;
- leading coefficient=0: a quadratic family can cease to be quadratic;
- physical inequality becomes equality: feasible domain boundary.
Always inspect the parameter value where the mathematical type changes.
Sensitivity Decision Tree
- What feature depends on the parameter?
- Can the feature be written explicitly? Vertex, root condition, period, optimum.
- Where can behaviour change type? Solve threshold equality.
- What happens below, at and above the threshold?
- Are any values inadmissible? Check domain/physical constraints.
- Can a graph family make the change visible?
Common Failure Modes
| Error | Cause | Repair |
|---|---|---|
| parameter treated as just another unknown | family meaning ignored | ask which feature it controls |
| threshold value omitted | regime transition not analysed | solve equality boundary explicitly |
| larger B in sin(Bx) said to increase period | inverse horizontal scaling missed | use period 2π/|B| |
| small rate change assumed to have small long-term effect | exponential sensitivity ignored | compare outputs after time |
| parameter change creates invalid model but is accepted | feasibility not rechecked | carry domain/physical constraints |
| one graph feature changes, others assumed to change too | invariants not identified | separate fixed and parameter-dependent features |
A 50-Minute Sensitivity Session
- 8 minutes: compare quadratic families as a,h,k vary.
- 8 minutes: map root-count regimes from a discriminant threshold.
- 8 minutes: compare exponential growth rates.
- 8 minutes: compare trig amplitude/period parameter changes.
- 8 minutes: derive an optimum as a function of a parameter.
- 10 minutes: identify threshold and feasibility changes in two mixed problems.
What Mastery Looks Like
- The learner interprets parameters as controllers of mathematical families.
- The learner identifies invariant and changing features.
- The learner finds threshold values where qualitative behaviour changes.
- The learner compares regimes below, at and above a threshold.
- The learner predicts graph, period, growth and optimum changes before full calculation.
- The learner recognises nonlinear sensitivity such as exponential compounding.
- The learner rechecks feasibility whenever a parameter changes.
Return to the Additional Mathematics Learning Hub
Additional Mathematics Learning Hub | Secondary 3–4 A-Math Guides
