Methods to fix values to two decimals

Does anyone know the most efficient way to fix values to two significant figures?

I am trying to proof a calculations which initially has the calculated values to two significant figures and then has the final values to 7 significant figures.

When researching I found a MatLab answer stating that you can not fix "binary floating point arithmetic" https://au.mathworks.com/matlabcentral/answers/16882-how-to-round-the-decimal-point

Does MatLab cater for fixed point instead of just floating points?

Answer 1

[Walter Roberson]

You can use the Fixed Point Toolbox.

You can also use the Symbolic Toolbox, but only if you convert the values into ratios, such as converting 0.23 to 23/100 (which sym() can do automatically for you.) If you use symbolic floating point numbers then there are situations where the calculation could be done in binary with the inexact representation problems you have heard about already.

When doing calculations to such limited precision, you need to think carefully about whether you want to round values or truncate values, and if you do want to truncate values, do you want to truncate towards 0 or do you want to truncate towards negative infinity ? For example if 0.236 is rounded towards 0 or negative infinity then that would become 0.23, but if -0.236 is rounded towards 0 that would be -0.23 whereas rounding towards negative infinity would be -0.24

[Jay]

Hi Walter,

Is there a way of using an if statement to specify the second decimal based on the third?

I.e. if 3rd decimal <5 then round down else round up?

[Walter Roberson]

Are you certain that is what you want? Consider -0.954 and -0.955 . For -0.954 the third decimal place < 5 so you want it to round down, which would be to -0.96 . For -0.955 the third decimal place is not < 5 so you want it to round up, which would be to -0.95 . This is a rounding scheme I have ever seen in use!

[Jay]

I see, thankyou.