What does the term “model-based” mean, and what are “signal-plus-noise” models?
The term “model-based” refers to the fact that estimates are derived by a statistical model rather than direct sampling. The approach used to estimate employment and unemployment in the census divisions, the states and the District of Columbia, and the eight substate areas is based on a signal-plus-noise model. This model postulates that the observed Current Population Survey estimate consists of a true, but unobserved, labor force value (the signal) plus noise that reflects the error arising from taking a probability sample rather than a complete census of the population. The modeling process separates the two to produce an estimate of the signal. For a more complete description, see Estimates for States.
