Abstract:
|
Many -- perhaps most -- preclinical animal experiments are poorly designed and executed. Two common problems that severely compromise the results are (1) treatment effects are confounded with biological and technical effects, and (2) the wrong biological entity is treated as the experimental unit, making the sample size artificially inflated (pseudoreplication). We discuss the criteria for determining the appropriate experimental unit and common sources of confounding in animal experiments such as sex, litter, time/order effects, and cages/pens/tanks. No new methods or theory are required to solve these problems, only the standard application of century-old statistical thinking.
|