The case for using Randomization in research

We all want to know that the work we do or the money we give makes a difference.  But how can we know?  For example, lets say we wanted to examine how an HIV educational program affected the sexual practices of teenagers.  One would naturally look to high schools who are teaching prevention education.  However, this is only a subset of the teenage population, since many teenagers aren’t in high school, such as those that have dropped out.  Collecting the sample in a random fashion is important to ensure that the information obtained is truly representative of the population.  In a randomized sample survey, the selection process is tailored so that each member of the population has an equal chance of being chosen to participate. This randomization guards against errors or biases. With a randomly selected sample, the differences between the two groups can be attributed to the intervention. 

In order for sample to be randomly selected, the population that is being sampled needs to be known.  An example of  this is in a population that is geographically specific, such as a district or a refugee camp. 

While this may be simplistic to say, a population cannot be sampled if it is not know.  An example of when randomization is not possible or feasible is seen when a subpopulation is being sampled that is imbedded in the general population and cannot be extracted or differentiated without significant cost.  This was recently seen in a survey of migrant workers in Beijing, where it was not possible from both cost and governmental approval to locate or distinguished migrants from nonmigrants within the city.