Post History

That depends on whether you are just building something for yourself, say a prototype, or if you are designing a product for mass distribution. In the later case, you are subject to probability laws, and the answer of Olin is just fine. 

In the former case, though, you may improve the performances by simply trying different resistor combinations (usually two or more equal resistors in series), and measuring the resistance with a Ohm meter. I did that several times in my lab to improve the accuracy of the resistance.

Doing so, it is usually better to try combination of resistors from different manufacturers, because resistors from the same manufacturer (or worse, from the same package) often possess the same bias.