-------- Original Message -------- Subject: choice of reference Date: Thu, 29 Apr 2010 17:42:38 -0400 From: Ben Wasserman <[email protected]> To: [email protected] CC: Eric Palkovacs <[email protected]> Hello All, I’m having trouble figuring out what to use as a reference figure in the following case. As part of a large study of guppies (Poecilia reticulata) we have photographs of some 1000 fish. For an additional 100 fish we have additional data: laboratory assays of feeding performance. At this stage I have only digitized the 100 fish, and used PLS to compare behavior to morphology within those 100 fish. The fish are from 12 different locations (high, medium, and low predation sites in each of 4 watersheds), but our subsample is only from the sites within a single watershed. My question is this, how can I project the results of this smaller study to say something about the potential performance capabilities of the fish in the larger study? If I had taken just a few linear measurements-say head depth, fish weight, and gape, then I could build a predictive model: Predicted performance = head depth + fish weight + gape But since the shape variables (be they aligned landmark coordinates or warps) are dependent on the reference one chooses to use, it is not as straight forward to me how to proceed. Right now this is my best idea, but I’m not sure if it is correct: redo the PLS on the shape scores calculated in the shape space defined by the consensus of all 1100 fish, and then projection becomes as straight forward as the above formula. Does an answer hinge on how representative these 100 individuals are of the shape space of all 1100? Thanks All! Ben Wasserman Ben.a.wasserman AT gmail DOT com -- For now, new message AND replies should be sent to: [email protected] /* Replies will be sent to the list. */ For more information visit http://www.morphometrics.org
