Study design is the planned method for assigning participants or observations to comparison groups so an analysis can evaluate a hypothesis. It connects the research question, data-gathering method, independent variable, dependent variable, and statistical test. section 3.0 section 3.2.4
A sound design begins with a clearly documented hypothesis and an understanding of the dataset. Exploratory data analysis checks distributions, outliers, missing values, duplicates, and summary statistics before results are prepared. section 2.1.3 section 3.0
Common comparison structures include:
- Between-subjects designs compare a dependent variable across mutually exclusive groups, such as treatment and control groups or geographic locations. section 3.2.4
- Within-subjects, or repeated-measures, designs expose every participant to each condition and compare repeated measurements, such as values before and after treatment. section 3.2.4
One hypothesis usually uses one study design and one statistical-test type, although a study evaluating multiple hypotheses can combine designs. The available designs are practical defaults rather than an exhaustive catalogue, so specialized questions may require other approaches. section 3.2.4
Website-layout experiment
An e-commerce team compares three website layouts to reduce abandoned carts. The analysis can examine geographic region, browser type, visit time, and previous visits as possible cohorts. If a layout performs differently from both the original and other test groups, the result requires careful interpretation and a recommended follow-up; extending the experiment can test whether performance changes as customers become familiar with the layout. section 3.2.5
