Brij Bhushan Vij wrote in USMA 21395:
>I think we can preresrve our hearts from 'beating or bleeding'. The SI
>manuals are clear on the subject that *ROUNDING* off must be done to keep
>the results at its most updated values - so if the result is 1, 2, 3, 4 at
>the last digit 'ignore it' and if it is FIVE (5) or above take it to the
>next higher place after the decimal point.!
> For example: 43.54 cm could be rounded to 44 cm (or 43.5 cm) and 43.67 cm
>could be rounded to 44 cm (or 43.7 cm). I suppose this is being done
>already.
>Brij Bhushan Vij
Those rules will produce a slight tendency for the mean of a set of figures
to increase after rounding. For example:
23.49 rounds to 23
23.51 rounds to 24
23.50 rounds to 24
The refined rule would round 23.50 to 24, but would also round 24.50 to 24.
The skewing of the mean after rounding is more pronounced after rounding
only one digit. Thus:
23.4 rounds to 23
23.5 rounds to 24
The refined rule is that
23.5 rounds to 24
but 24.5 also rounds to 24
I have not be able to write a simple computer subroutine to handle this
more strict rule, so I have ignored the problem.
Joseph B.Reid
17 Glebe Road West
Toronto M5P 1C8 Tel. 416 486-6071
