[EMAIL PROTECTED] (Patricia Bacon) wrote in message news:<[EMAIL PROTECTED]>... > Dear readers of EdStat, > > I have a data set of the number of tourists who came > to Brazil > from 1990 to 2000. I have 10 different nationalities > of tourists > and I also have the type of accommodation they were > in. I also > have the different cities in Brazil and the number of > tourists > for each one, also by years and even months, by kind > of > accomodation and by nationality. > > I would like to extract the information of those data > in order > to find if there are patterns among the tourists, the > evolution > of them, the differences in the kind of accomodation, > if it has changed through the years and if there are > seasonal patterns. > > I checked cluster analysis, but since I have ten > variables corresponding to the years, I don't know if > this is correct. > > I was told that a correspondence analysis should be > appropriate (for the tourists and the kind of > accomodation), > but since I have the variable year and month I don't > know how to deal > with it and maybe there are more appropriate > techniques. > In order to show the evolution, I think I should > extract the > information for each nationality as a time series, but > I don't know > if I could also show the evolution for several > nationalities altogether. > As I have the data by months, ARIMA models should fit > (I guess). > > If someone knows how I can deal with this data, the > Kind of > techniques I should use and if there are any > references of related > things, I would appreciate it so much. > > Thank you very much in advance. > > Patricia > > ___________________________________________________ > Yahoo! Messenger - Nueva versi�n GRATIS > Super Webcam, voz, caritas animadas, y m�s... > http://messenger.yahoo.es > . > . > ================================================================= > Instructions for joining and leaving this list, remarks about the > problem of INAPPROPRIATE MESSAGES, and archives are available at: > . http://jse.stat.ncsu.edu/ . > =================================================================
Patricia, ARIMA modelling can be done for individual series in a class and then done globally for all the series in the class. This leads directly to a formal test of the hypothesis of a common model across all series. If one rejects the hypothesis of equality then one can then find what contrast caused the difference ala Schefee or Duncan type conclusions. Dave Reilly AUTOMATIC FORECASTING SYSTEMS http:/www.autobox.com http:/www.autobox.com/freef.exe for a FREEWARE version of AUTOBOX . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
