Hi I'm looking for information on outlier detection (possibly using nonparametric methods) methods for multivariate data fields that are irregularly sampled in space and time. The particular application we are working on entails forming spatial grids from the scattered data for each month. All data values that fall in the month in the defined grid box are to be averaged. The identifcation of outliers before averaging the data is desired. It is felt that information on space and time neighbors should be useful for classifying individual data points as outliers or not, and for identifying "features" that are extremes in space and/or time. We have some ideas on things we would like to try. However, I would like suggestions and pointers to existing literature where this problem has been addressed. I'll be happy to summarize useful responses to this group. Thanks in advance Upmanu
