Rounding, see if you want to keep integer or several decimal places, and then directly round to less than 4 according to the last digit, and add 1 to the previous digit greater than or equal to 5. For example: keep an integer, 2.4 ≈ 2; 2.5≈3。 Keep one decimal place, 2.24 ≈ 2.2; 2.25≈2.3

If the probability of 0 ~ 9 is equal, then the total error of this method is the smallest for a large number of retained data, which is probably why we take this method as the basic retention method.

For example:

1.15+1.25+1.35+1.45 = 5.2, if one decimal place is rounded:

1.2+ 1.3+ 1.4+ 1.5=5.4

According to the calculation of "four houses in six and fifty pairs",1.2+1.2+1.4+0.4 = 5.2, the rounding result can better reflect the actual result.

In C#, the result of int. Tostring ("F2") is calculated according to the rule of four families, and six families are divided into 50% pairs. Rounding in other calculation software, such as Matlab, is mostly handled in this way.