Many statisticians prefer randomization-based analysis of data that was generated by well-defined randomization procedures.

The randomization procedures used in the experiment have been criticized for not randomizing satisfactorily.

However, many of the included studies received lower quality scores because they failed to describe the randomization procedure and participant withdrawals or drop-outs from the study.

The protocol specifies a randomization procedure for the experiment and specifies the primary data-analysis, particularly in hypothesis testing.

An ideal randomization procedure would achieve the following goals:

A good randomization procedure will be unpredictable so that investigators cannot guess the next subject's group assignment based on prior treatment assignments.

One of the best ways to ensure independence is to properly follow rigorous randomization procedures.

The treatment groups were determined by a computer-generated randomization procedure.

The randomization procedure was performed three times and statistics were calculated on the average output of these runs.

One should check the success of the randomization procedure, for instance by checking whether background and substantive variables are equally distributed within and across groups.