Thursday, November 22, 2012

Why Do We Randomize in Experiments?

Curious people experiment.  We want answers we can't find on Google.  Unfortunately, without some training, experimental results may not be valid.

One of the first things to know about experimenting is the need to randomize. First, randomization helps avoid the pitfalls of bias. Bias is the tendency for the data to all move one direction due to something other than what we are interested in investigating.

Second, statistical analysis makes some assumptions. We assume all errors will gather about an average and approximately cancel to zero. Without that assumption, many statistical tools are useless. So we randomize to justify making that assumption.

Design of experiments is a complex subject. But learning some simple principles will help you get started with everyday experiments that answer YOUR questions.

Happy experimenting!

Mrs. Fields

123112

No comments:

Post a Comment