Comparing Trends in Unmatched Timed Data/b/mscom/archive/2007/06/06/comparing-trends-in-unmatched-timed-data.aspxIntroduction – Maybe You Have Been There Too
Ok, this could be another pretty dry topic, but here we go anyway. More programmer talk, which can put normal people to sleep. J A while ago I came up with a programmatic solution to a challenge I had facedWe are the operations team that runs the Microsoft.com sites.7.x ProductionRE: Comparing Trends in Unmatched Timed Datahttp://blogs.technet.com/b/mscom/archive/2007/06/06/comparing-trends-in-unmatched-timed-data.aspxThu, 07 Jun 2007 09:32:08 GMTd5e57398-b9ef-4490-9955-07cbb4e4a80d:bb7d75d9-317e-4e0c-85cf-123ec5389dbfAnonymous0<p>I question the method of doubling the values in stock B to compare trend to stock A. If you double the values of stock B, you also double the rate of change between those values. dy/dx(x)=1, but dy/dx(2x)=2 for all x. Doubling the values to make rate of change (derivative) comparison is not a valid comparison; stock B in your last graph above is twice as curvy as it should be.</p>
<p>Instead, add a constant to move the graph of one stream up or down relative to another. </p>
<p>-john</p><img src="http://blogs.technet.com/aggbug.aspx?PostID=1164154&AppID=4713&AppType=Weblog&ContentType=0" width="1" height="1">