## Significant figures and rounding electricity dance moms episode

What is the difference between the two population numbers stated above? The first one expresses a quantity that cannot be __known exactly__ — that is, it carries with it a degree of uncertainty. It is quite possible that the last census yielded precisely 157,872 records, and that this might be the population of the city for legal purposes, but it is surely not the true population. To better reflect this fact, one might list the population (in an atlas, for example) as 157,900 or even 158,000. These two quantities have been rounded off to four and three significant figures, respectively, and the have the following meanings:

• 1579 00 (the __significant digits__ are underlined here) implies that the population is believed to be within the range of about 157850 to about 157950. In other words, the population is 157900±50. The plus-or-minus 50 appended to this number means that we consider the *absolute uncertainty* of the population measurement to be 50 (50) = 100. We can also say that the **relative uncertainty** is 100/157900, which we can also express as 1 part in 1579, or 1/1579 = 0.000633, or about 0.06 percent.

Which of these two values we would report as the population will depend on the degree of confidence we have in the original census figure; if the census was completed last week, we might round to four *significant digits*, but if it was a year or so ago, rounding to three places might be a more prudent choice. In a case such as this, there is no really objective way of choosing between the two alternatives.

This illustrates an important point: the concept of *significant digits* has less to do with mathematics than with our confidence in a measurement. This confidence can often be expressed numerically (for example, the height of a liquid in a measuring tube can be read to ±0.05 cm), but when it cannot, as in our population example, we must depend on our personal experience and judgement.

So, what is a significant digit? According to the usual definition, it is all the numerals in a measured quantity (counting from the left) whose values are considered as **known exactly**, plus one more whose value could be one more or one less:

Suppose, for example, that we measure the weight of an object as 3.28 g on a balance believed to be accurate to within ±0.05 gram. The resulting value of 3.28±.05 gram tells us that the true weight of the object could be anywhere between 3.23 g and 3.33 g. The *absolute uncertainty* here is 0.1 g (±0.05 g), and the **relative uncertainty** is 1 part in 32.8, or about 3 percent.

How many significant digits should there be in the reported measurement? Since only the left most 3 in 3.28 is certain, you would probably elect to round the value to 3.3 g. So far, so good. But what is someone else supposed to make of this figure when they see it in your report? The value 3.3 g suggests an implied uncertainty of 3.3±0.05 g, meaning that the true value is likely between 3.25 g and 3.35 g. This range is 0.02 g below that associated with the original measurement, and so rounding off has introduced a bias of this amount into the result. Since this is less than half of the ±0.05 g uncertainty in the weighing, it is not a very serious matter in itself. However, if several values that were rounded in this way are combined in a calculation, the rounding-off errors could become significant.