Multiple stratification when one of the stratifying variables is time

Stratified random sampling is one of the most widely used sampling techniques. In practice, in many surveys, it is possible to stratify the population with respect to a number of characteristics. Multiple stratification has been used in practice for some time and a number of alternative designs have been proposed but there appears to be little attempt to apply these techniques to the case when one of the stratifying variables is time. The aim of this thesis is (i) to consider existing and possible alternative 2-way stratification designs and (ii) to investigate alternative procedures by which temporal stratification can be continued to give sequences of interviewers work loads across strata. The relative efficiencies of such procedures depend on the relative magnitudes of strata effects, time effects, and their interactions. Data collected in a bus survey is used to estimate such effects.Some of the existing 2-way stratification designs are extended to more general conditions, corrected or improved. We also consider a new 2-way stratification design and show that it always yields a solution.We consider 2 new 2-way stratification designs when one of the stratifying variables is time. It is shown that the procedures which take into account the serial correlation which might be present due to the time variable are at least as precise as ordinary 2-way stratification designs.We also consider 2 new 3-way stratification designs when one of the stratifying variables is time. Again the designs which take into account the serial correlation arc found to be more efficient.