Hierarchical randomization can reduce type I errors, solve the imbalance of important covariates and improve the confidence of small sample experiments. According to the characteristics of the research object, that is, some factors that may produce mixed effects (such as age, gender, race, education level, living conditions, etc. ), the subjects were stratified first, and then randomly assigned to A (experimental group) and B (control group) at each layer.
Commonly used randomization methods include simple randomization (such as lottery and random grouping), block randomization (grouping objects with similar characteristics into the same block and randomly grouping them within the block), and stratified randomization (stratification factors can be age, gender, disease degree, etc. In order to make the factors that have a serious impact on the results as balanced as possible in each group).
Hierarchical random operation steps:
1. Select the stratification factor to determine the stratification level.
2. Determine the number of research cases at each level, which is usually an integer multiple of the intervention measures. In order to ensure the random effect, it is better to be more than 5 times.
3. Then each research object in the layer generates a random number.
4. Allocate according to the defined allocation principle, such as even numbers belong to group A and odd numbers belong to group B. ..
For example:
10 male and female subjects were randomly divided into two groups according to their age. To make the number of subjects in each layer 10, it is necessary to complete all the grouping of the two layers (with the age of 60 as the boundary). Random number sorting, specifying 1-5 as group A and 6- 10 as group B.