The reasoning for this study was that if the sample pairs displayed enough redundancy, then analyzing a single sample per week could result in substantial savings. Three questions were examined: (1) to what extent are the weekly pairs correlated with each other; (2) how well do they each correlate with the self-reports of cocaine use; and (3) whether group comparisons differ based on the sample used and in comparison with an analysis using all samples. Twice weekly urine samples from 71 subjects collected over 11 weeks were examined for quantitative urine benzoylecgonine concentration (BE) levels. Agreement between pairs of samples was estimated for both quantitative and qualitative measures of BE. Results appear to be contradictory: within-week samples were not correlated with each other but both samples were equally correlated with measures of self-report and both individually, and when combined led to the same conclusion regarding treatment effects. This inconsistency lies in the different focus each analysis brings to the data. As expected, quantitative urine concentrations of cocaine metabolite levels vary extensively with study participants within each week. The findings are consistent with the known pharmacokinetics of cocaine. This is important in tracking an individual’s cocaine use as opposed to the behavior of the group as a whole as measured by mean levels. If primary interest is on the effects of a treatment on the group of study participants as a whole, the findings indicate that a single sample taken from a weekly pair of samples can lead to similar conclusions as using both samples to estimate the group’s behavior. At least two samples per week may be necessary for an accurate representation of cocaine use in an individual. 2 figures, 3 tables, 15 references