Seminar
Alan Dorfman
The Two Sample Problem
It sometimes happens that two separate samples from a population, having perhaps quite distinct
designs and mode of sampling, contribute information on the same variable of interest, and it
becomes an important question how to combine the data from the two samples. Examples are the
National Health Interview Survey (NHIS) and the Behaviorial Risk Factor Surveillance System (BRFFS),
both contributing data on, for example, smoking habits, and the Occupational Employment Statistics
survey (OES) and the National Compensation Survey (NCS), both contributing information on occupational
wages.
After giving some background on OES/NCS (both carried out by the Bureau of Labor Statistics), we discuss
options that have been considered for combining the data and achieving unified estimation.
The problem here considered may be regarded as a special case of the more general problem of combining
data from disparate sources, for example, an observational study and clinical trial that each provide
evidence on the same treatment.