Continuous distributions

S2 Topic 3, spec point by spec point (Continuous distributions). Each spec point has the method, worked examples written the way the mark scheme wants them, how the marks go and the mistakes that cost the most marks.

Spec3.1The continuous uniform (rectangular) distribution

Use the continuous uniform distribution, including deriving its mean, variance and cumulative distribution function.

The idea

Every value between \(a\) and \(b\) is equally likely, so the pdf is flat:

\[f(x) = \frac{1}{b - a}, \quad a \leqslant x \leqslant b \qquad E(X) = \frac{a + b}{2} \qquad \text{Var}(X) = \frac{(b - a)^2}{12} \qquad F(x) = \frac{x - a}{b - a}\]

Probabilities are just lengths: \(P(c < X < d) = \frac{d - c}{b - a}\).

Deriving the variance (you may be asked)

\(E(X^2) = \displaystyle\int_a^b\frac{x^2}{b - a}\,dx = \frac{b^3 - a^3}{3(b - a)} = \frac{a^2 + ab + b^2}{3}\).

\(\text{Var}(X) = \dfrac{a^2 + ab + b^2}{3} - \dfrac{(a + b)^2}{4} = \dfrac{a^2 - 2ab + b^2}{12} = \dfrac{(b - a)^2}{12}\).

Typical uses

Rounding errors, a random cut point, arrival times within an interval. If a condition is an inequality in \(X\) (e.g. a quadratic), solve it first and then find the length of the region.

cutting a string

A 40 cm string is cut at a random point. \(L\) is the length of the longer piece. Find the distribution of \(L\) and the probability the longer piece is more than three times the shorter.

\(L \sim U[20, 40]\).

\(L > 3(40 - L) \Rightarrow L > 30\), so \(P = \frac{40 - 30}{20} = \frac12\).

an unknown end point

\(W \sim U[-6, a]\) and \(\text{Var}(W) = 27\). Show \(a = 12\).

\(\dfrac{(a + 6)^2}{12} = 27 \Rightarrow (a + 6)^2 = 324 \Rightarrow a + 6 = 18 \Rightarrow a = 12\) (positive, since \(a > -6\)).

  • B1 for "uniform" or "rectangular", dB1 for the interval (WST02A Jan 2026 Q5a).
  • M1 for the variance equation, A1* cso; verifying by substituting 12 is not allowed for a "show that" (WST02 Jan 2024 Q5a).
  • Inequalities: M1 form, M1 solve, dM1 add the right regions, A1 (WST02 Jan 2021 Q5d).
  • \(E(3 - 2X) = 3 - 2E(X) = 3 - a - b\) (WST02 Jun 2017 Q7a).

Often asked as: very common (108 questions). Probabilities; mean and variance; unknown end points; derive \(E(X)\) or \(\text{Var}(X)\); functions of a uniform variable.

\(\text{Var}(X) = \frac{(b - a)^2}{2}\) or \(\frac{b - a}{12}\).

Spec3.2The Normal approximation to the binomial and Poisson

Use the Normal distribution as an approximation to the binomial and Poisson distributions, with a continuity correction.

When

Exact Approximation Conditions
\(B(n, p)\) \(N(np, np(1 - p))\) \(n\) large, \(p\) close to 0.5 (e.g. \(np > 5\) and \(n(1 - p) > 5\))
\(\text{Po}(\lambda)\) \(N(\lambda, \lambda)\) \(\lambda\) large (e.g. \(\lambda > 10\))

The continuity correction

The binomial and Poisson are discrete (whole numbers); the Normal is continuous. Each whole number \(k\) is represented by the interval from \(k - 0.5\) to \(k + 0.5\).

Discrete Continuous (with correction)
\(P(X \leqslant 47)\) \(P(Y < 47.5)\)
\(P(X < 47)\) \(P(Y < 46.5)\)
\(P(X \geqslant 35)\) \(P(Y > 34.5)\)
\(P(X > 35)\) \(P(Y > 35.5)\)
\(P(X = 40)\) \(P(39.5 < Y < 40.5)\)

Then standardise as in S1: \(Z = \frac{Y - \mu}{\sigma}\), using \(\sigma = \sqrt{\text{variance}}\).

binomial

\(J \sim B(150, 0.4)\). Use a Normal approximation to find \(P(J \leqslant 47)\).

\(\mu = 60\), \(\sigma^2 = 36\), so \(J \approx N(60, 6^2)\).

\(P(J \leqslant 47) \approx P(Y < 47.5) = P\left(Z < \frac{47.5 - 60}{6}\right) = P(Z < -2.08) = 1 - 0.9812 = 0.0188\).

Poisson

\(X \sim \text{Po}(30)\). Approximate \(P(X \geqslant 35)\).

\(X \approx N(30, 30)\). \(P(Y > 34.5) = P\left(Z > \frac{4.5}{\sqrt{30}}\right) = P(Z > 0.82) = 1 - 0.7939 = 0.206\).

  • B1 (or M1 A1) for the Normal with the right mean and variance, M1 for standardising (with a continuity correction value), M1 for the right continuity correction, A1 for the probability (WST02 May 2024 Q3f).
  • Unknown \(n\): standardise with the continuity correction in terms of \(n\), equate to a \(z\) value, and solve the quadratic in \(\sqrt{n}\) (WST02 Jan 2021 Q3).

Often asked as: very common (161 questions, including S1 Normal questions filed here). Approximate a probability; use it inside a hypothesis test (4.6); choose between Poisson and Normal approximations with a reason.

No continuity correction, or the wrong direction.

Using the variance instead of the standard deviation.