Fundamentals of experimental design: lessons from beyond the textbook world

Authors: Casler, Michael

Submitted to: Agronomy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/4/2013
Publication Date: N/A
Citation: N/A

Interpretive Summary: We often think of experimental designs as analogous to recipes in a cookbook. We look for something that we like and frequently return to those that have become our long-standing favorites. We can easily become complacent, favoring the tried-and-true designs (or recipes) over those that contain unknown or untried ingredients or those that are simply too weird for our tastes. Instead, I prefer to think of experimental designs as a creative series of decisions that are meant to solve one or more problems. These problems may be real or imagined - we may have direct evidence of a past or current problem or we may simply want insurance against future potential problems. The most significant manifestation of a "problem" is a failure to detect differences between experimental treatments. Six tenets of experimental design - randomization, degree of replication, form of replication, size of experimental unit, level of blocking, and size/shape of blocks - can be used creatively, intelligently, and consciously to solve both real and perceived problems in comparative experiments. This paper will be of value to any biological researcher.

Technical Abstract: We often think of experimental designs as analogous to recipes in a cookbook. We look for something that we like and frequently return to those that have become our long-standing favorites. We can easily become complacent, favoring the tried-and-true designs (or recipes) over those that contain unknown or untried ingredients or those that are simply too weird for our tastes. Instead, I prefer to think of experimental designs as a creative series of decisions that are meant to solve one or more problems. These problems may be real or imagined - we may have direct evidence of a past or current problem or we may simply want insurance against future potential problems. The most significant manifestation of a "problem" is unsatisfactory p-values that prevent us from developing inferences about treatment differences. Six tenets of experimental design - randomization, degree of replication, form of replication, size of experimental unit, level of blocking, and size/shape of blocks - can be used creatively, intelligently, and consciously to solve both real and perceived problems in comparative experiments.