Talk:Secular variation

From Wikipedia, the free encyclopedia

The phrase "secular variation" is not a suitable term for an encyclopedia article.

I disagree. I'm not able to expand this because I don't really know time series, but I have no doubt that this could be expanded far beyond a dictionary definition, and I expect anyone who realizes that time series has been an area of active research for many decades would agree with me. Michael Hardy 20:20, 22 February 2007 (UTC)

Well, I don't want to argue with you, Michael. If you want to delete the "PROD" tag, go ahead. I'll just leave the article alone if you do that. I do have about 20 years' experience with time series data, and in fact I often worked to eliminate the secular trend from a time series, so that I could model the residuals better. But I would put a description of such techniques under "spectral analysis", or "regression analysis", or some heading like that for the purposes of an encyclopedia. The "secular variation" is often just the straight line that best fits (by least squares) the time series data, and the residuals from that regression sometimes show some sort of periodic or quasi-periodic behavior. DavidCBryant 21:25, 22 February 2007 (UTC)

I thought about this a little longer, and you might be right, Michael. They say that familiarity breeds contempt. Maybe I got so used to thinking of a secular trend as a triviality (when I worked with that sort of data) that I can't really be objective now. I mean, the really interesting part of many time series is what's left over after any secular trend has been eliminated. So that's what I think is important. I guess it might not look like that to somebody else. Weird, eh? DavidCBryant 21:44, 22 February 2007 (UTC)