Planning an Experiment

(adapted by Duane Deardorff from Experimentation by D.C. Baird)

Careful planning of experiments can save time and money. Even more significant is the fact that a poorly designed experiment can yield completely useless results. Below are some basic steps for planning a scientific experiment.

1) Identify the system and model — Clearly defining the system under investigation is necessary for maintaining focus on the purpose of the experiment. Quite often an experiment becomes so cluttered with ancillary equipment or other details that we can loose sight of the fundamentals. Clearly identify the model that explains the system and the limitations or assumptions of the model.

2) Choose the variables — Identify the dependent variable that is desired as the result of the experiment, along with all possible factors that can influence this result. Select which of these input variables will be the primary independent variable and how the other factors can be held constant or varied in discrete steps.

3) Rectify the equation — Attempt to describe the model of the system mathematically and linearize the model if possible. It is usually best to arrange the linear equation so that the desired result can be found from the slope rather than an intercept, because intercepts are often subject to systematic errors that may be difficult to identify and eliminate.

4) Choose the ranges for the variables — Before taking measurements, decide on a reasonable range for the independent variables. A range of at least a factor of 10 usually gives satisfactory coverage for analyzing the behavior of the system, but a wider range that includes extreme cases is best for understanding the limits of the system. However, care must be taken to not exceed instrument limits that might damage equipment. These limits can usually be determined from instrument ratings and a few cautious trial measurements. Once identified, these limits should be carefully observed throughout the experiment to avoid delays caused by unnecessary replacement of expensive equipment.

5) Consider the overall precision and timeline for the experiment — A general sense of the overall experimental uncertainty and the amount of time expected to obtain the result is needed as a guide for designing the experiment. Once a realistic goal is established, it is possible to ensure that all of the measurements are of sufficient precision to contribute usefully to the final result and that time and effort is not wasted making overly precise measurements. An analysis of the sources of error can also be used to identify the factors that most significantly limit the precision of the result.

6) Perform a quick test (pilot study) — Before spending months or years on a project, it makes sense to test the experimental conditions to determine if the project is even feasible as planned. There is no greater frustration than to spend large amounts of time and effort preparing to obtain a result that is completely useless or unobtainable with the existing experimental design.