Performing scientific investigations of poison exposure in the lab, we have an experimental design with a set of reference and a treated samples (say A-samples and B-samples). We ad some chemical and measure a response at several time intervals after the exposure � and at each point in time A and B measurements are made in triplicate (response could be concentration of a metabolite, number of living cells, a color or the like). We then make an XY (time/response) plot and fit a curve (could be an exponential decay for example). As we measure in triplicate at each time there is variation on each measurement point on both A and B values. Though, in spite of fluctuations and decreasing response for both the A and the B curve it seems that the A response curve is higher than the B curve "most of the time"� How do we make a statistical test of this? It should takes into account the decay (response levels are not constant over time) and preferably the variation in the triplets observations?
In a reply to a similar question (posted by Don Taylor to newsgroup sci.stat.math 1997/05/26 ) it was suggested that in AUTOBOX the problem could be treated as a 'POOLED CROSS-SECTIONAL TIME SERIES' � however, it was not clear how to treat this problem in general (we do not have AUTOBOX). Does anyone have a hint or reference how to solve this problem? Thanks in advance . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
