Chappell R
Department of Statistics, University of Wisconsin, Madison 53706.
Stat Med. 1991 Dec;10(12):1945-58. doi: 10.1002/sim.4780101209.
A technique is presented which provides guidance on the spacing of follow-up waves in a multiwave study. Only information from the baseline wave is needed, as well as rough parameter estimates for the survival distribution. The computations use the expected Fisher information; a new method for its calculation is given. An analysis of the effects of simplifying assumptions, such as presuming that a distributional parameter is known, shows that gross discrepancies may result. Data from the Massachusetts Health Care Panel Study (MHCPS) are used for a demonstration. The MHCPS example shows that this technique can prevent large inefficiencies in multiwave experimental design.
