Rounding Off

What information does rounding preserve?

Rounding replaces a number with a nearby value expressed at a specified precision. It preserves scale and approximate magnitude while discarding some detail. The requested place value determines which detail is retained: nearest whole number, nearest ten, one decimal place and three significant figures are different instructions. A rounded answer is not an exact value, and its usefulness depends on whether the discarded detail can affect the decision being made.

State the rounding convention when a value lies exactly halfway between two candidates. School-style exercises commonly round a positive halfway value upward, while some computational systems use a ties-to-even rule. These conventions agree for most numbers but differ at certain ties. A well-posed practice question specifies the rule where the answer depends on it, rather than silently treating one implementation as universal.

How are decimal places different from significant figures?

Decimal places count positions after the decimal point. Rounding 12.346 to two decimal places gives 12.35 under the usual nearest-value rule. Significant figures count meaningful digits beginning at the first nonzero digit. The number 0.004786 to two significant figures becomes 0.0048. Its leading zeros locate the decimal scale; they are not the first significant figures.

The distinction matters when quantities differ greatly in size. Two decimal places can retain substantial detail for 127.43 but erase a small quantity such as 0.004786 to 0.00. Significant figures express a different idea of precision. Do not infer the intended uncertainty of a measurement merely from an ambiguous trailing zero in a whole number. If precision matters, the problem should specify it or use a notation that makes the intended digits clear.

What bounds follow from a rounded measurement?

Suppose a positive length is reported as 8.4 metres after rounding to the nearest tenth, with halfway values rounded upward. Its original value lies from 8.35 metres inclusive to 8.45 metres exclusive. Values just below 8.35 would round to 8.3; the upper halfway value rounds to 8.5. The interval describes all original values consistent with the report, not a claim that the actual length is uniformly distributed within it.

For rounding to the nearest unit q, the maximum absolute rounding error is at most q/2 under nearest rounding. An exact endpoint's inclusion depends on the tie rule. This error bound is more informative than assuming that the displayed value is exact. When several rounded measurements enter a calculation, each contributes uncertainty. A neat-looking final decimal cannot restore information that the original reports omitted.

Why can rounding twice give a different result?

Consider 4.449 under the usual positive half-up convention. Direct rounding to one decimal place gives 4.4 because the hundredths digit is 4. If the number is first rounded to two decimal places, it becomes 4.45; rounding that result to one decimal then gives 4.5. The first rounding has moved the number onto a later decision boundary.

Keep enough working digits and round once at the stage requested. If a problem explicitly requires rounding intermediate values, follow that instruction and distinguish the result from direct final rounding. The discrepancy is not evidence that arithmetic has become inconsistent; it comes from performing different sequences of approximation. In multiple-choice work, unnecessary intermediate rounding can shift an answer into a nearby distractor even when the underlying method is correct.

When is an estimate sufficient to choose an option?

Suppose a calculation asks for 19.8 multiplied by 5.1 and the options are 10, 100, 1,000 and 10,000. Estimating 20 multiplied by 5 gives about 100 and clearly separates the choices. If the options are 99.98, 100.98, 101.98 and 102.98, the same estimate cannot decide the answer. Exact multiplication gives 100.98. The required precision depends partly on how closely the alternatives are spaced.

Use bounding when possible. Since 19.8 lies below 20 and 5.1 lies above 5, replacing both factors does not automatically create an upper or a lower bound. One change pushes the product up and the other down. To establish a bound, choose replacements with known effects in the same direction, or calculate an explicit error estimate. Approximation without tracking direction is useful for scale but weaker for close comparisons.

How do absolute and relative error differ?

An absolute error of 0.5 unit has different significance for a value near 2 than for one near 2,000. Relative error divides the absolute error by the relevant exact or stated reference value. For an original value of 40 rounded or approximated to 42, the absolute error is 2 and the relative error is 2/40 = 5 per cent. Dividing by the approximation would answer a different ratio.

For a sum, errors can accumulate. If three values are each rounded to the nearest whole number, the total absolute error can be as large as 1.5 units, though actual errors may partly cancel. Cancellation should not be assumed without evidence. For multiplication and division, examine the scale and sensitivity of the expression rather than adding raw errors with different units. A denominator close to zero can make a small change particularly consequential.

What is a dependable final checking routine?

Write the requested precision beside the answer space before starting. Preserve exact fractions or adequate digits during the main calculation, then round once using the specified convention. Check the order of magnitude, units and whether a halfway case occurs. If the input itself is rounded, report a justified range or acknowledge the resulting uncertainty when the question asks what can be determined.

The purpose of rounding is controlled simplification. It can save time, expose an implausible answer and communicate sensible precision, but it should not disguise missing information. In practice, distinguish three statements: the exact result, a rounded representation of it, and a broad estimate used for elimination. Keeping those statements separate prevents a shortcut from being mistaken for a proof of a closely spaced numerical answer.