> Could you give some names and numbers? I can follow your criticism. Okay: My object of interest is a urban sewer system. The system consists of single pipes or reaches (my "objects"). For every reach I've got a number of independent metrical variables like construction date, diameter, coverage as well as non-metrical like material or profile type.
A sewer system can be divided into sub-catchments, similar to river watersheds. A sub-catchment contains a number of reaches (my "grouped objects"), a reach belongs to only one subcatchment. For every sub-catchment I've got a infiltration rate of parasite water, my dependent variable. Due to the available measuring methods I know this rate only integrated for the whole sub-catchment. Connecting the length-specific rate to the reaches is nonsense. Calculating the correlation between infiltration rate and mean values independent variables is possible, but then I neglected the number of reaches (which differs), the distribution of independent variables .... So what can I do not losing to much information? You see, with your "bag of vegetables" you were on the right way. > Especially from the Subject line, I wonder if has to do with > "aggregation" ... but ... Nearly. My hypothesis is that reaches with similar characteristics have similar infiltration rates. Thus, I'm looking for aggregation or clustering. The problem is that beside my independent variables I have to consider the ramification structur (I hope it's the right word here) and the measuring methods. Torsten . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
