Stratification is, broadly speaking, the act of dividing a population into subgroups. In the context of research, stratification is a concept that spans across two stages of the research process:

• Recruitment, Data Collection: Stratification typically comes into play during participant randomization (i.e., when participants are assigned to a study group). It is a way of guaranteeing that previously defined subgroups—generally based on characteristics related to the research question—are represented in each study group. For example, you might stratify randomization by sex, which will mean that each of your study groups will have the same ratio of females to males. 
• Data Analysis: Stratification then carries through into the data analysis stage as you try to understand the impact of the subgroups on the data. In many cases, the factors defining the subgroups will be confounders—something that complicates the relationship between the variables being studied—so the stratification helps control and understand the impact those confounders have on the results.

OMF’s Melbourne ME/CFS Collaboration, for example, will utilize stratification in the analysis of their study called Serial Paediatrics Omics Tracking in Myalgic Encephalomyelitis (SPOT-ME). Their goal with stratification in this case is to understand how biological sex impacts the relationship between markers in the blood and symptoms in the participants. For more information about this project and the stratified analysis, you can read the SPOT-ME protocol paper in BMJ Open.