On Mon, 17 May 2004 16:22:40 +0200, "Rafal" <[EMAIL PROTECTED]> wrote:
> Hello > > I have a regular network of sample plots distributed evenly over the area > under research. On each plot 6 samples were taken. There is big variability > between averages from the plots even if they are closest neighbors. I was > suggested that I should resample my data by aggregating neighboring plots > and creating one plot in the middle of the cluster which will give me a plot > with 24 observations. What is the most strange to me is that I was suggested > that I should use one plot for 4 neighboring clusters to avoid bias in > space. I have a very mixed feelings about using one plot in four > neighboring plots. Of course it smoothes the data, but I think that it > happens not because the number of observations increased but because the > same neighboring plots are taken to neighboring clusters. If two plots next to each other are greatly different, compared to the consistency of the 6 measures taken on each plot, that result seems to demonstrate that something is effective about the treatment that differentiates the plots. If the two plots had the same treatment and yet differ that way, then it has to indicate that the locations make a difference, and the interaction has to serve as the error term for testing. >From the little you have implied about the purpose, I don't see any justification for averaging or smoothing. i also don't see the meaning of 'resampling' in the subject line. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
