In article <[EMAIL PROTECTED]>, James Choe <[EMAIL PROTECTED]> wrote:
>Thanks Radford for your clear explanation. I think I am more >interested in directional hypothesis, then. Would Kruskal-Wallis >suffice in that case? If not, what sort of test should I be using? (I >have monthly time-series county level data from 1980 - 1997). I am >learning stats off of a book, but I am wasn't sure which test is >appropriate for a case I described. It probably looses a bit of information, but the simplest approach I can think of would be to start by taking yearly averages for all counties in one state and all counties in the other state. Perhaps a weighted average, weighted by size of county would be appropriate, though as I mentioned, the whole idea of comparing precipitation by state (a totally artificial entity) seems rather strange. Then take the differences. You're left with a time series of differences in precipitation for 1980-1997. You then need to estimate the autocorrelations, and use that to adjust the standard error for the estimated mean difference. I think Diggle's book on time series covers this. However, 18 years may not be enough data. Also, there may be a trend (or long-lag autocorrelation) over the 18 years, which would make the results meaningless. I suggest you think more about what you're really trying to accomplish. Radford Neal ---------------------------------------------------------------------------- Radford M. Neal [EMAIL PROTECTED] Dept. of Statistics and Dept. of Computer Science [EMAIL PROTECTED] University of Toronto http://www.cs.utoronto.ca/~radford ---------------------------------------------------------------------------- . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
