Singapore School Mathematics Operating Manual · Chapter 27
Not every mathematical rule behaves the same way everywhere.
A fare may change after two hours. A tax rate may change after a threshold. A graph may change direction at a stationary point. A quadratic may move from two real roots to one repeated root to no real roots as a parameter crosses a critical value. A feasible region may suddenly become empty. A rounding decision may change once an uncertainty interval crosses a boundary.
These are regime changes: points where one mathematical description stops being enough and another becomes active.
This chapter develops the operating discipline for thresholds and critical values: identify the controlling parameter, locate the boundary, classify the regions on each side, test the boundary itself, and state exactly where each conclusion holds.
1. A threshold is a value where decision status changes
If admission requires score ≥ 75, then 75 is a threshold.
74.9 fails while 75 passes under an exact continuous model.
The interesting mathematics often lies near the boundary because small changes can alter classification.
2. Inclusive and exclusive thresholds are different
x ≥ 5 includes the boundary.
x > 5 excludes it.
This one symbol can change whether a critical case belongs to the feasible set.
3. Piecewise functions encode regime changes directly
Suppose C(t)=3t for 0≤t≤2, and C(t)=6+5(t−2) for t>2.
The pricing rule changes at t=2.
To evaluate C(1.5), use the first branch. To evaluate C(4), use the second.
The threshold determines which mathematical law applies.
4. Boundary continuity is a separate question
At t=2, the first branch gives 6.
The second branch would also approach 6 from above, so the model is continuous there.
A regime change does not necessarily mean a jump in value; it may mean a jump in slope or rule.
5. Some thresholds create discontinuities
A fixed entry fee applied once usage exceeds zero can create a jump between zero usage and positive usage.
Likewise, discrete eligibility rules can change status abruptly at a cutoff.
The type of boundary matters: value jump, slope change, feasibility change or classification change.
6. Quadratic discriminants create three regimes
For ax²+bx+c=0 with a≠0:
Δ=b²−4ac > 0 gives two distinct real roots.
Δ=0 gives one repeated real root.
Δ<0 gives no real roots.
Zero is the critical discriminant threshold separating root regimes.
7. Parameter questions often ask for where the regime changes
Consider x²−4x+k=0.
Δ=16−4k.
Two distinct real roots require k<4.
A repeated root occurs at k=4.
No real roots occur for k>4.
The parameter k crosses a structural threshold at 4.
8. Monotonicity changes at critical points
For f(x)=x²−6x, f′(x)=2x−6.
The derivative changes sign at x=3.
The function decreases for x<3 and increases for x>3.
x=3 is therefore a regime boundary in behaviour.
9. Stationary points are not automatically extrema
A derivative can equal zero without the sign changing.
For f(x)=x³, f′(0)=0 but the function is increasing on both sides.
The boundary must be classified by nearby behaviour, not merely detected.
10. Absolute value creates a sign threshold
|x|=x for x≥0 and |x|=−x for x<0.
The rule changes at x=0.
Many absolute-value derivations are therefore regime problems disguised as one symbol.
11. Rational expressions can have forbidden thresholds
For f(x)=1/(x−2), x=2 is not a valid point.
The domain splits into x<2 and x>2.
The function’s behaviour on the two sides cannot be connected through the excluded point.
12. Logarithms create domain boundaries
log(x−3) requires x>3.
x=3 is not a valid input; values below it are outside the real domain.
Domain boundaries can therefore be more fundamental than ordinary thresholds: one side may not belong to the model at all.
13. Square roots create inclusive domain thresholds
√(x−3) requires x≥3.
Unlike logarithms, the boundary x=3 is allowed.
This illustrates why boundary inclusion must be derived from the operation, not memorised by topic label.
14. Discrete systems turn continuous thresholds into whole-number decisions
A bus holds 40 passengers.
At 40 passengers, one bus is enough.
At 41, two are needed.
The operational output jumps even though the input rises by only one.
Ceiling-type rules create stepwise regimes.
15. Floor and ceiling functions are threshold machines
The floor function stays constant across intervals and jumps at integers.
The ceiling function does likewise.
These functions model capacities, batches, complete groups and minimum resource counts.
16. Rounding creates half-unit thresholds
To round to the nearest integer, boundaries occur at values ending in 0.5 under the usual convention.
A value of 7.49 rounds to 7 while 7.51 rounds to 8.
Near the threshold, small measurement changes alter the reported integer.
17. Uncertainty intervals can straddle a threshold
If a true value lies in [9.8,10.2] and the decision threshold is 10, the classification is not guaranteed.
Some compatible values lie below; others lie above.
The correct conclusion is conditional or uncertain, not a forced yes/no.
18. Guaranteed decisions require the entire interval on one side
If the value lies in [9.1,9.7] and threshold is 10, it is guaranteed below.
If it lies in [10.3,10.8], it is guaranteed above.
This is the strongest form of threshold reasoning under uncertainty.
19. Percentage thresholds depend on the correct base
A 20% increase threshold should be measured relative to the original value.
Using the final value as base can move a quantity into the wrong classification.
Threshold reasoning is only as sound as the underlying ratio model.
20. Probability thresholds support classification but not certainty
If an event has probability above 0.5, it is more likely than not under the model.
This does not mean the event must occur.
Probability thresholds classify likelihood, not deterministic outcomes.
21. Correlation cutoffs are conventions unless mathematically defined
Calling a correlation “strong” above some numerical cutoff is often a reporting convention, not a universal theorem.
Threshold labels should distinguish mathematical necessity from contextual convention.
22. Feasible regions can appear or disappear at critical parameters
Suppose x≥0 and x≤k.
If k≥0, feasible values exist.
If k<0, no value satisfies both.
The parameter threshold k=0 separates non-empty and empty feasible sets.
23. Geometry has existence thresholds
Three positive lengths form a non-degenerate triangle only if the largest is strictly less than the sum of the other two.
Equality produces a degenerate straight-line case.
Greater-than produces impossibility.
The triangle inequality therefore defines a structural regime boundary.
24. Circle intersection has three regimes
Two circles may intersect in two points, one tangent point, or no points depending on centre distance and radii.
Tangency is the critical boundary between intersection and separation.
Geometry often contains exactly this pattern: two solutions, one repeated configuration, none.
25. Optimisation often changes owner at a boundary
A global maximum on a closed interval may occur at an interior stationary point for one parameter range and at an endpoint for another.
Parameter changes can therefore shift which candidate class controls the optimum.
26. Model rules can change because the real mechanism changes
A transport fare can move from one rate to another after a distance threshold.
A production system can add another machine after demand crosses capacity.
A school model may encode such mechanisms with piecewise equations.
27. Regime changes are not necessarily smooth
Some are continuous but change slope.
Some jump in value.
Some change only logical status.
Some remove feasibility entirely.
Classify the type of change rather than treating every threshold as the same phenomenon.
28. A critical value should be tested separately
When solving parameter inequalities, do not classify only “less than” and “greater than”.
The equality case often has special behaviour, as with Δ=0 for repeated roots.
The threshold itself is frequently the most informative case.
29. Backward reasoning can find the threshold directly
If the target is “at least 100”, solve the equality boundary first.
For 20+5n≥100, solve 20+5n=100 to get n=16, then use monotonicity to classify n≥16.
The boundary is often easier to find than evaluating many cases.
30. Scale and thresholds interact
If an answer is orders of magnitude away from a threshold, exact refinement may be unnecessary.
If it is extremely close, small errors and assumptions matter.
Decision effort should increase near critical boundaries.
31. Sensitivity matters most near thresholds
A one-unit perturbation is irrelevant if the result is 1,000 units away from a boundary.
The same perturbation can flip a decision when the result lies 0.1 units from it.
This links regime reasoning to the next chapter on Conditioning and Ill-Posedness.
32. A practical threshold audit
Ask:
What parameter controls the decision? Where does the rule or status change? Is the boundary included? Is there a separate equality case? Does the mathematical formula change, or only the interpretation? Could uncertainty cross the threshold? Is the conclusion stable on both sides?
33. Independent practice
1. Classify x²−4x+k=0 by number of real roots as k varies.
2. For |x|, state the two formula regimes.
3. State the real domain of √(x−5).
4. State the real domain of log(x−5).
5. A bus holds 30 passengers. How many buses are needed for 30, 31 and 60 passengers?
6. A value lies in [4.7,5.3] and the threshold is 5. Is the classification guaranteed?
7. A value lies in [4.1,4.6] and threshold is 5. What is guaranteed?
8. Explain why Δ=0 deserves a separate case.
9. Give one example of a threshold that changes slope but not value.
10. For 3n+2≥20, find the critical equality value and classify n.
11. Why should equality at a triangle-inequality boundary not be treated as an ordinary triangle?
12. Why does uncertainty matter more near a decision boundary?
34. Worked answers
1. Δ=16−4k. Two roots for k<4, repeated root at k=4, no real roots for k>4.
2. |x|=−x for x<0 and |x|=x for x≥0.
3. x≥5.
4. x>5.
5. 1, 2 and 2 buses respectively.
6. No. The interval crosses 5.
7. The value is guaranteed below 5.
8. It corresponds to a repeated root and is structurally different from both positive- and negative-discriminant regimes.
9. A continuous piecewise tariff whose rate increases after a time threshold.
10. Equality at n=6. Hence n≥6.
11. Equality makes the points collinear, producing a degenerate case rather than a non-degenerate triangle.
12. Small perturbations can flip the decision when the result is close to the boundary.
35. Continue through Batch 07
Use Inverse Problems when thresholds must be reconstructed from observed outcomes, Symmetry and Case Reduction when several regimes are equivalent, and Conditioning and Ill-Posedness when tiny input uncertainty can move a result across a critical value.
Return to the BTT Mathematics Hub for Batch 07.