Try something that doesn't have a systematic seasonal component to it. For example, the temperature on a particular day (October 25, say) over several years. Or the difference between the daily temperature and the mean temperature for that day (the information available would likely be the average temperature for the day, and it might have been averaged over a number of days, perhaps a month, which would make this residual-from- overall-average difficult to observe). Or the difference between the maximum temperature for the day and the mean temperature (this might be more likely to be recorded somewhere, and has the additional advantage of having a mean different from zero; the minimum temperature would work as well; if you used both, you could then ask whether the minimum was as far below the mean as the maximum was above the mean, on the average).
For any of these quasi-residual measures, you could lump together data from many days. (Don't lump together data from different locations, though. The daily variation in temperature is MUCH greater in, say, Melbourne (Australia) than it is in Washington, e.g., and you'd end up (possibly unknowingly!) with a mixture problem.) On Thu, 24 Oct 2002, Stan Brown wrote: > My class is starting normal distributions today, and I asked them to > name some distributions and say whether they were normal. > > One student suggested seasons, such as high temperatures. Those are > more nearly cyclical, but then I got to wondering if there are any > seasonal or climate phenomena that are normally distributed. I can't > think of any, but I wonder if I'm overlooking an obvious example. > > Any suggestions? ----------------------------------------------------------------------- Donald F. Burrill [EMAIL PROTECTED] 56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816 [was: 184 Nashua Road, Bedford, NH 03110 (603) 471-7128] . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
