Dimensional analysis

Dimensional analysis is the technique of carrying units alongside numbers through a calculation and treating them as algebraic objects. Units multiply, divide, and cancel; if your final answer is supposed to be a distance and the units work out to seconds, you know you’ve made an error before you even check the numbers.

Worked example. Converting 60 mph to meters per second:

60 mi/hr × (1609.344 m / 1 mi) × (1 hr / 3600 s)
= 60 × 1609.344 / 3600 m/s
= 26.8224 m/s

Notice how the “mi” and “hr” units cancel diagonally. The result has units m/s — which is what was asked for. Any other unit in the answer would indicate a math error.

This is more than a teaching technique. NASA’s 1999 Mars Climate Orbiter loss happened because Lockheed Martin’s ground software output pound-seconds and the spacecraft’s software expected newton-seconds. The conversion factor wasn’t applied; the resulting trajectory commands sent the orbiter into the Martian atmosphere. $125 million plus the mission.

For human calculations, dimensional analysis catches roughly:

• Forgotten conversions (miles to kilometres, hours to seconds).
• Inverted ratios (multiplied when should have divided, or vice versa).
• Conflated quantities (treating mass as weight; treating volume as flow).

The rule of thumb: every step of every calculation should have units attached. The numerical answer is right only if the unit answer is also right.

Worked example 2 — drug dosing

A paediatric IV order reads “morphine 0.1 mg/kg every 4 hours; child weighs 18 kg; stock solution 2 mg/mL.” Work the units out fully: dose mg = 0.1 mg/kg × 18 kg = 1.8 mg. Volume = 1.8 mg ÷ 2 mg/mL = 0.9 mL. Notice how the mg cancel diagonally and the answer comes out in mL — exactly the unit you measure with a syringe. If anywhere in the calculation you ended up with a unit like “mg²/mL” or “kg/mg”, you’d know to stop and recheck before drawing the syringe.

When and why it matters

Dimensional analysis matters wherever quantities with different units interact — engineering, physics, medicine dosing, finance (return rates × time = total return), and cooking (mass / volume = density). The classic medical-error case: a paediatric drug dosed by “mg per kg of body weight” gets a decimal-point misread, and a 10× overdose ships out the pharmacy door because the nurse didn’t cross-check “does this volume make sense for a 12 kg toddler?” A 30-second dimensional sanity check (dose volume × concentration = total mg; total mg ÷ body weight = mg/kg; compare to prescribed mg/kg) catches the error every time. In software, type systems like Haskell’s dimensional or F#’s units-of-measure encode dimensional analysis at the compiler level, making unit mismatches a build error rather than a runtime catastrophe. The 1983 Gimli Glider — a Boeing 767 that ran out of fuel mid-flight because ground crew confused pounds and kilograms during refuelling — is the aviation case study taught in every flight-safety course. Reference: NIST — The International System of Units (SI).

A useful tactic in engineering review: write every constant in a formula with explicit units, even when the formula is being copied from a textbook. The gravitational constant g is not just “9.81” — it’s 9.81 m/s². The speed of light is not just “3×10⁸” — it’s 3×10⁸ m/s. When the formula is applied, the units propagate and any missing factor (a kilometres-to-metres conversion, for instance) immediately fails the dimensional check before the wrong number ever reaches a spec sheet.

Buckingham π theorem — the deep version of the same idea: Edgar Buckingham’s 1914 result formalises dimensional analysis as a structural property of physical equations. Any meaningful physical equation can be reduced to a relationship between dimensionless groupings of variables — the Reynolds number in fluid dynamics (ρvL/μ), the Mach number (v/c), the strain (ΔL/L), the drag coefficient. The theorem says the number of dimensionless groups equals the number of variables minus the number of independent fundamental dimensions. It’s the basis for the famous “scaling argument” in physics — predicting that a 1:100 scale model wind-tunnel test transfers to full-scale behaviour requires matching the Reynolds number, not just the geometry. Reference: NIST SP 811 — Guide for the Use of the SI.