On 6 Aug 2003 06:10:40 -0700, [EMAIL PROTECTED] (John G.) wrote: > Hi everyone > > I need some help comparing curves generated from an experiment > examining ecto-parasite loads on fish. > > A number of individuals were infected and the number of parasites > counted on each fish five times over the next 30 days. The fish showed > three basic patterns of response to the initial infection. The first > group showed an exponential-like growth of parasites until the end of > the experiment (we call these susceptible). The second group showed an > initial increase in numbers until around the mid-point after which the > numbers dropped, sometimes to zero but usually to low levels (the > curve looks like a hill or dome) (resistant type 1). The final group > seemed to retain a low load throughout the experiment, neither > increasing or decreasing (resistant type 2).
After reading the whole post, my first thought was that there are two dimensions, which could be described in various ways. The "final level" and the "average level" would seem to cover the three trajectories. (You might be working with the logarithms, but that is a minor issue.) If the numbers fall out as neatly as they are described, then you might be like an alternative, single number that would be the slope of the final 3 (of the five) measures. This has the unusual feature, though, of possibly being disordinal -- I don't know the subject, but it seems to me that this ordering raises intriguing questions. (For instance, could the lack-of-change owe to the environment, say, rather than to the specimen? Or, if the parasites were a cold virus, the resistance would demonstrate a previously-gained immunity.) - The "Big increase" would be the worst group, - The "Big decrease" would be the intermediate group, since it *did* get infected. - The "Little change" (last three times, remember) would be the group that avoided infection: presumably the best immune-response. > > I want to be able to do two things with this data. Firstly, I would > like to be able to split the fish into the three groups mentioned > above using stats, not charts. Secondly, I would like to generate some > value/s which I could use as a guide to ‘strength of > response’. In this case I would imagine getting something like a > high value for resistant type 1 as they show a definite and strong > response, this might be followed by a medium value for resistant type > 2 as although they did not reduce the parasite load, it was prevented > from increasing to high levels, and finally a low value for the > susceptible group as they showed no response at all. > > Please help. I have played with fitting and comparing various curves, > putting the data through PCA etc but am having no luck achieving my > two aims. Orthogonal components would probably emerge as: Mean; Linear fit across time; Quadratic fit. Your groups sound as if they could be separable by means alone, so I wonder that you don't mention why that obvious answer does not work. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html "Taxes are the price we pay for civilization." Justice Holmes. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
