News for hackers. Stuff that matters.

A comment on Data scientists need to learn about significant digits. Parent comment

Re:Data scientists need to learn about significant digits by SAI_Peregrinus on Monday December 25, 2023 @06:16PM (#38767256)
If you have a material sample with measured length of 1cm ±0.5cm, and a measured mass of 3g±0.5g, then reporting 3.3333333333g/cm for mass per unit length is just lying. It's between 0.14 and 0.6, so "about 0.3g/cm" is not lying, whereas "0.333333333333g/cm" is.
[ Parent | on HN ]
Re:Data scientists need to learn about significant digits by mitthrowaway2 on Monday December 25, 2023 @06:32PM (#38767362)
What do you think about 0.333 ± 0.222 g/cm?
[ Parent | on HN ]
Re:Data scientists need to learn about significant digits by SAI_Peregrinus on Monday December 25, 2023 @09:35PM (#38768452)
0.3 (-0.2+0.3) g/cm would be more correct. You've still got 3 significant figures when the inputs only had 1. Just don't lie, it's not that hard.
[ Parent | on HN ]
Re:Data scientists need to learn about significant digits by mitthrowaway2 on Monday December 25, 2023 @09:40PM (#38768474)
Interesting definition of a "lie". I don't think anything about that is inaccurate or untruthful. At best it could be called verbose, but yours is actually no shorter. Confidence intervals supersede significant figures in all respects, and now you've introduced a small but unnecessary bias that depends on the peculiarities of base 10.
[ Parent | on HN ]