On Mon, 6 Jan 2003 [EMAIL PROTECTED] wrote: > As is often the case, please tell us what you want really to do (in your > substantive application) rather than for a vague statistical procedure, > and we may be able to point you to appropriate tools.
Thanks - happy to oblige. Sorry if my original question was a bit vague. I have a set of narrative texts and want to carry out a "narrative pattern analysis" for each text. I divide each text into consecutive segments of N (usually ca.200) running words each and count for each segment the overall frequency of words that populate the conceptual category that I'm interested in, giving me a frequency count of, say, "emotion" for each segment. I then want to see whether this category shows a significant linear or nonlinear development through the narrative, or whether its occurrence is random. However, other people who have done this have suggested that autocorrelation within the series can amplify or diminish actual differences between segments and therefore that autocorrelations should be removed, even if they are not statistically significant. I know how to fit a basic linear or nonlinear curve, but I don't know how to go about getting rid of these autocorrelations. Many thanks, Andrew Wilson ______________________________________________ [EMAIL PROTECTED] mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-help
