Q. Of course we can't lose sight of original units of coefficient (for instance, in regression above, we have to know that B1 coefficient is measured as ‘dollars-per-barrel-of-gasoline/dollars-per-barrel-of-crude-oil'. But we can also change units into ‘standard errors' so that we can ask, of beta estimate, how many standard errors it is away from zero. If it is not very far from zero n we might ignore it; if it is far away from zero n we might consider it important. But how far is ‘far'?