Random treatment assignment with unequal treatment fractions and dealing with misfits.
Download all files from this repository and add them to your personal ado path. The command is also available on the SSC repository.
randtreat is a Stata program that performs random treatment assignment. It can handle an arbitrary number of treatments and unequal treatment fractions, which are common in real-world randomized control trials. Stratified randomization can be achieved by optionally specifying a variable list that defines multiple strata. It also provides several methods to deal with misfits, a practical issue that arises in treatment assignment whenever observations can't be neatly distributed among treatments. The command performs all tasks in a way that marks misfit observations and provides several methods to deal with those misfits.
When run, it creates a new variable named treatment that encodes the treatment allocation. The default number of treatments is two (a control group and a treatment group), but more can be specified with the mult(integer) option. Also, unequal fractions of treatments may be specified using the unequal(fractions) option. Random assignment with all these options can be done within each strata defined by the unique combinations of values of varlist. The data will be sorted by varlist and treatment, unless the sortpreserve option is issued. The seed can be set with the setseed(#) option, so the random assignment can be replicated.
When the number of observations in a given stratum is not a multiple of the number of even treatments, each stratum will have misfits (i.e. units that can't be neatly distributed among the treatments). If unequal treatments are specified, then the problem arises whenever the number of observations in a given stratum is not a multiple of the least common multiple (LCM) of the fractions' denominators. When the command is run, it will display the number of misfits that the current setup yields. By specifying misfits(method), one can choose how to deal with those misfits.
Please refer to the help file for further detail.
One of the first to discuss the 'misfits' issue were Bruhn and McKenzie (2011), in a World Bank Blog post. A generalization of the problem and details of the Stata implementation can be found in Carril, 2016.