Sequential rerandomization

Area of Science:

• Statistics
• Experimental Design
• Clinical Trials

Background:

• Traditional rerandomization assigns all units at once.
• Sequential unit enrollment is common in practice, especially for rare diseases.
• Existing methods lack a framework for sequential rerandomization.

Purpose of the Study:

• To develop a mathematical framework for sequential rerandomization designs.
• To formulate an adaptive procedure for balancing treatment/control assignments.
• To compare the covariate balance of sequential versus one-time rerandomization.

Main Methods:

• Developed a mathematical framework for sequential rerandomization.
• Formulated an adaptive procedure using Mahalanobis distance for imbalance.
• Analyzed expected covariate balance under sequential enrollment in groups.

Main Results:

• Sequential rerandomization provides a framework for grouped unit enrollment.
• The adaptive procedure balances treatment/control assignments using Mahalanobis distance.
• Sequential rerandomization achieves better expected covariate balance than one-time rerandomization with equal rerandomizations.

Conclusions:

• Sequential rerandomization is a viable and effective alternative to one-time rerandomization.
• This method improves covariate balance in practical experimental settings with sequential enrollment.
• The proposed framework supports adaptive balancing for continuous or binary covariates.