Originally posted by Blue 2008 and 2009 had more area on than 2007 had.
So what? Are you back to arguing trends?
Quote: Where are you using SD instead of SEM?
Do you mean
why? For one, because we would like to know the variance of the actual sample sets, not the variance of the mean estimates of some hypothetical large population the samples were extracted from, but regardless, if you want to calculate the SEMs, just divide the stdevs by the square roots of 22 and 3, respectively. It makes no difference to the statistics.
Quote: 2007 is the lowest data set that they have on record (which isn't the same thing as the lowest in history). Even then, they aren't separated by a lot.
They are separated by well over 3 standard deviations at the trough, which is usually enough to conclude that a data point is with very high probability outside of a distribution (in this case, we have three consecutive data points outside of the distribution). And indeed, actual statistical analysis shows that the probability that the 2007-2009 data come from a 1979-2000-like distribution is, effectively, nil.
Quote: What's the rationale for the data set breaking at 2000 rather than 1989 or 90?
Not sure, other than bigger samples are better, and values were fairly stable throughout those 22 years and seem to have dipped severely only recently.
But anyway, I can't hold your hand through more of these phony statistical concerns. I have shown you where the data are, analyze them whichever way you want and let us know what you find. I don't feel like going back to step one of your four-step dance.