# 3 Basic study design principles

- In all aspects of conducting a study we are trying to ensure we can make valid inference. In this sense much of application of statistics is concerned with ensuring valid inference.
- In many types of studies, we often try to ensure that there is an element of randomness in the way in which we derive a sample. This is often done to avoid introducing bias, an example might be to help make the sample representative of the target population.
- Randomisation is a tool we use in experiments to control for sources of unwanted variation between treatment or exposure groups. This is critical for ensuring comparability between groups and hence for allowing causality to be attributed to any association we observe.
- The concept of randomisation is completely different to the idea of a random sample; randomisation refers to the stochastic allocation of the observational units to the exposure, random sampling refers to the stochastic sampling of observational units from the population into a study.

- Understanding the sampling process, ie, how we going from population to sample, is of fundamental importance to drawing sensible conclusions from the results of a study.
- The way that observations are sampled directly affects the types of methods that should be used for statistical inference. For example, studies that recruit from families or clusters of individuals that share characteristics eg, sampling by school, require special methods to ensure valid
*statistical inference* to account for this shared variation. We do not cover these methods in this course but you should be aware of when they may be needed.
- The way in which we extract a sample from a population can be influenced largely by our research question and the approach used is a key feature that characterises the study design, we pick up these ideas again in the study design theme.