When selecting a

value to use in the forward finite difference approximation of the first derivative, one should always:

Group of answer choices

Use the machine epsilon value to achieve maximum level of precision.

Use $10^-8$ regardless of which precision is being used $($double or single precision$).$

Use the smallest number possible because truncation error will be less.

Use a value that best balances the opposing effects of truncation and round$-$off errors for the desired level of precision