Experienced readers critically evaluate experimental design before drawing conclusions about a study.

The reliability of a study is largely determined by the overall experimental design.

A case study intensely describes one subject or a small group of subjects. Scientists performing case studies can only draw conclusions about the subjects examined. Case studies often lead to the formation of hypotheses that can be tested by more rigorous experimental designs.

A cross-sectional study is the observation of a defined population at a single point in time. Scientists do not actively manipulate experimental variables in this design, rather they look for connections among variables. Cross-sectional studies are useful for establishing correlation between two variables, but they usually do not establish causation.

A longitudinal study involves data collection over a defined period of time. Scientists establish experimental treatments and examine the effects of changing a particular variable. This design allows the researcher to measure changes in variables over time in response to an experimental treatment and establish causation.

Research question: Do low carbohydrate diets promote weight loss?

Case study: A scientist places two overweight patients on a low carbohydrate diet and measures weight. If these patients lose weight, there is suggestive evidence that a low carbohydrate diet may contribute to weight loss in some individuals.

Cross-sectional study: A scientist records the diets and weights of 500 individuals. The weights of those who eat a low carbohydrate diet are compared to those eat other diets. This study could establish a correlation between low carbohydrate diets and lower body weights. However, it can not establish that low carbohydrate diets cause lower body weights, because many factors have not been controlled in this study design.

Longitudinal study: A scientist places 20 subjects on a low-carbohydrate diet and compares their weight to 20 control subjects on a normal diet. Because this study establishes an experimental treatment (the low carbohydrate diet) and controls other variables, it could support the hypothesis that low carbohydrate diets cause lower body weights.

The sample size is the number of subjects or individuals studied. In scientific literature, the letter "N" (or "n") is used to designate sample size. In experiments, scientists usually work with a sample of subjects rather than the full population. Good experimental design must include a sample of subjects that is representative of the greater population being studied.

Scientific hypotheses are tested by contrasting a "control group" with an "experimental group". In experimental groups, the variable under study is manipulated. In control groups, the variable under study is not manipulated. Control groups are important because they serve as a reference point for the experimental groups. In some studies, subjects are studied twice, once with the experimental manipulation, and once with a control manipulation. In this case, the subject's control measurements can be compared to their own experimental measurements.

Scientists examining human subjects often use placebos in their control treatments. A placebo is a substance or treatment that replicates the experimental treatment but lacks the active component. Placebos give the control subjects the same expectation as the experimental subjects and thus enable scientists to control for psychological effects.

Scientists often need to replicate an experiment several times to establish a clear relationship between control and experimental groups. Each performance of the experiment is called a replicate. The number of replicates that a scientist chooses to perform is important; too few replicates could lead to inconclusive results while too many replicates is a waste of resources.

Click here for a review of the methods section.