MATSEC AM Applied Mathematics common mistakes
MATSEC AM Applied Mathematics · Understand the traps and how to avoid them.
Treating acceleration as constant in a problem where the force (and hence acceleration) varies with time or position, e.g. applying v = u + at to a particle whose acceleration is given as a function of x, or vice versa.
Candidates memorise the five suvat equations and pattern-match on 'kinematics' without first checking whether acceleration is uniform. The equations are only valid for constant acceleration, and the habit of reaching for them is reinforced by most textbook drills using uniform acceleration.
What to do instead
Before writing any suvat equation, explicitly state the acceleration and ask 'is this constant?'. If a = f(t), integrate with respect to t (a = dv/dt) to get v, then integrate again for s. If a = f(x) or a = f(v), use a = v dv/dx or a = dv/dt with the chain rule. Write the chosen differential relation at the top of the working so the examiner sees the method.
Resolving forces without a clear, consistently drawn force diagram, and silently omitting a force (usually weight, the normal reaction, or a tension in a connecting string) so the equilibrium or Newton's-second-law equations are incomplete.
Candidates try to write equations straight from the prose, keeping the force picture 'in the head'. Under exam pressure the weight of a mass on an inclined plane, or the reaction at a hinge, is easily forgotten because it is not mentioned as an applied force.
What to do instead
Draw a large free-body diagram for each body separately, marking every contact force and the weight, and label angles. Tick off each force in the diagram against the equation. For connected particles draw one diagram per particle and state that tension is the same throughout an inextensible string over a smooth pulley.
Taking moments about a point and getting the perpendicular distance wrong, for example using the slant length instead of the perpendicular distance for a force at an angle, or forgetting to include the weight of a uniform rod acting at its midpoint.
The moment formula 'force × distance' is over-simplified in candidates' minds; the perpendicular-distance condition is dropped. Uniform rods and laminae are treated as massless or their centre of mass is misplaced.
What to do instead
Always write moment = F × d where d is explicitly defined as the perpendicular distance from the pivot to the line of action. If a force acts at angle θ to a rod, use F sin θ × length (or the perpendicular component). Mark the centre of mass of any uniform body at its geometric centre and include its weight as a separate force in the moment equation.
Applying the friction law F = μR blindly when the body is not on the point of slipping, or using the wrong normal reaction on an inclined plane (using mg instead of mg cos θ).
The limiting-friction case is the one practised most, so candidates assume limiting friction in every equilibrium problem. The reaction on an incline is often taken as the full weight because the geometry is not resolved carefully.
What to do instead
State whether the body is in limiting equilibrium (then F = μR) or in equilibrium with friction below the limit (then F is found from the equilibrium equations and only checked against μR). Always resolve perpendicular to the plane first to get R; on a plane inclined at θ, R = mg cos θ when no other perpendicular force acts.
Misapplying conservation of momentum or the impulse–momentum principle by ignoring direction signs, or treating an inelastic collision as if kinetic energy were conserved.
Candidates set up momentum equations in mixed directions on a line without a sign convention, so a momentum in the negative direction is added rather than subtracted. The phrase 'collision' triggers a vague memory that energy is conserved, when only momentum is conserved in all collisions and kinetic energy only in perfectly elastic ones.
What to do instead
Draw a before/after diagram with an arrow marking the positive direction and write every velocity with its sign. State the principle used: momentum conserved for all collisions; kinetic energy conserved only if the question says elastic or the coefficient of restitution e = 1. Use impulse = change in momentum = ∫F dt for force–time problems, and Newton's law of restitution e = (separation speed)/(approach speed) for collisions.
Using the work–energy principle or power formula with the wrong quantity, e.g. equating work done by a driving force to the change in kinetic energy while ignoring work done against friction and against gravity, or using P = Fv with F as the total force rather than the driving force.
The energy equation is treated as 'KE before = KE after' rather than a full accounting of all work terms. In power problems the traction force and the resistive force are conflated.
What to do instead
Write the full work–energy statement: work done by driving force = change in KE + change in PE + work done against friction (or use the energy equation with all terms listed). For power, use P = Fv where F is the driving/tractive force, and at constant speed set F equal to the total resistance; when accelerating, use F − R = ma and P = Fv simultaneously.
Adding probabilities of non-mutually-exclusive events without subtracting the intersection, or confusing P(A|B) with P(B|A) in conditional and Bayes-type questions.
The addition rule and the conditional formula look similar and candidates apply them mechanically. Tree diagrams and two-way tables are under-used, so the direction of conditioning is lost.
What to do instead
For 'or' questions use P(A∪B) = P(A) + P(B) − P(A∩B) and check whether the events can overlap. For conditional questions draw a tree or two-way table and read P(A|B) as 'restrict to the B branch'; use P(A|B) = P(A∩B)/P(B). Test independence by checking P(A∩B) = P(A)P(B) rather than assuming it.
Using the binomial distribution when trials are not independent or the probability is not constant (sampling without replacement), or using the normal approximation to the binomial without the continuity correction or without checking np and n(1−p) are large enough.
Binomial is the default 'count of successes' model in candidates' minds, and the conditions (fixed n, constant p, independence) are not checked. The continuity correction is a small detail that is easily skipped once the standardisation formula is memorised.
What to do instead
Before choosing a distribution, list its conditions: binomial needs fixed n, two outcomes, constant p, independence; if sampling is without replacement from a small population use the hypergeometric or treat as binomial only when the population is large. When approximating, state np > 5 and n(1−p) > 5 (or the syllabus threshold), and apply the ±0.5 continuity correction before standardising.
Computing the mean and standard deviation from raw data or grouped frequency tables with the wrong formula, for example using n instead of n−1 for a sample standard deviation, or using class midpoints without weighting by frequency.
The distinction between population and sample standard deviation is blurred, and grouped data is treated as if each midpoint occurred once. Calculator statistics modes are used without understanding which formula is being applied.
What to do instead
Write the formula explicitly before substituting: for a sample use s² = Σ(x − x̄)²/(n − 1); for grouped data use Σfx/Σf and Σf(x − x̄)²/Σf (or the n−1 version as required). For grouped data always multiply by frequency and use midpoints, and state the assumed midpoint for open-ended classes.
Interpreting a correlation coefficient or regression line as proof of causation, extrapolating a regression line far outside the data range, or misreading the regression coefficients (using the x-on-y line to predict y from x).
Candidates learn 'r close to 1 means strong relationship' and slide into causal language. The two regression lines are not distinguished, and extrapolation is done because the arithmetic is easy.
What to do instead
State that correlation measures linear association only and does not imply causation; mention confounding variables. Check the data range and refuse to extrapolate beyond it. Identify which line is required: to predict y from x use the y-on-x line (minimising vertical deviations), and quote r² as the proportion of variation explained where relevant.
Carrying out a hypothesis test with the wrong null/alternative direction (one-tailed vs two-tailed), or drawing a conclusion that says 'the null hypothesis is accepted' or misstates significance in context.
The mechanics of the test (test statistic, critical value, compare) are rehearsed, but the wording of the conclusion and the choice of tail are treated as afterthoughts. Candidates also confuse 'not rejecting H0' with 'proving H0 true'.
What to do instead
Write H0 and H1 explicitly with the correct inequality before computing anything; use a two-tailed test unless the question indicates a direction. Compare the test statistic with the critical value (or p-value with α) and conclude in context: 'there is sufficient/insufficient evidence at the 5% level to reject H0 that ...'. Never say H0 is accepted; say it is not rejected.
Setting up a mathematical model without defining variables and assumptions, then failing to interpret the answer back in the real context or to comment on the model's limitations.
Candidates treat modelling questions as pure algebra and skip the 'define, assume, solve, interpret, evaluate' cycle that the applied-context questions reward. Assumptions such as 'the particle is a point', 'air resistance is negligible', 'the string is light and inextensible' are left implicit.
What to do instead
Adopt a fixed five-step structure: (1) define variables with units, (2) state assumptions, (3) set up equations, (4) solve, (5) interpret the result in context and comment on whether the assumptions are reasonable. In mechanics name the standard assumptions explicitly; in statistics state the sampling assumptions. Reserve one or two sentences at the end for evaluating the model.