Stratified random sampling divides the population of N units into subpopulations of N N NL units, according to differences in some defining characteristic such as household income. Random samples are then taken within each stratum. For example, a household activity and travel behavior survey in Portland, Oregon, used 10 strata: residents in four counties, park-n-ride users, and five subgroups in the largest and most urban county. These five subgroups were residents in the following areas: (1) urban, good pedestrian environment, land-use mix, transit availability; (2) urban, bad pedestrian environment, transit availability; (3) urban good pedestrian environment, transit availability; (4) light rail corridor; and (5) other residents of the county [Cambridge Systematics, 1996a]