This paper proposes to develop an estimate called a national projection from the NCSS data. It then evaluates the statistical properties of this estimate. Such an estimation procedure is applicable, under certain assumptions, to study designs where the first stage sampling units are chosen purposively. National projections for accident statistics from the NCSS data are produced. The sensitivity of the national projections on the choice of first stage units and on the models used in the estimation procedure are investigated.
All statistics in this paper are calculated from the first fifteen months of the NCSS data, January 1, 1977 to March 31, 1978.