-------- Original Message -------- Subject: Reconstructing data using Morpheus et al. Date: Sat, 30 Jan 2010 14:30:04 +0100 From: [email protected] To: morphmet <[email protected]> Hi morphometricians, I am a PhD candidate studying craniometric diversity in Upper Palaeolithic and Mesolithic populations. As I am sure many of you can appreciate, most of my datasets contain missing values. I plan to reconstruct these data using a combination of reflected relabelling and multiple multivariate regressions. This is my hypothetical workflow for reconstructing data, followed by some questions regarding my protocol. I've included Morpheus menu commands in square brackets: *Reflected relabelling* 1) Perform a GPA on the original data for individual specimen and its reflected relabelled counterpart. [Points > Superimposition > GPA (=GLS)] 2) Compute mean configuration (creating a symmetric form). [Points > Superimposition > Save > Means (scaled)] 3) Restore the centroid size, location and orientation of the mean configuration to return it to the coordinate space of the original specimen. [*see note below] *Multiple multivariate regressions* 4) Perform a GPA on complete dataset. [Points > Superimposition > GPA (=GLS)] 5) Restore scale, so that landmarks of each specimen are aligned but size differences are preserved. [Points > Superimposition > Restore > Scale] 6) Estimate missing values using multiple multivariate regressions. My first concern is regarding the validity of my protocol. If there are any obvious flaws, I would be grateful if someone could point them out. In Morpheus et al. is there a command that would allow me to retrieve the centroid size value from a GPA analysis? The only way I have been able to do this is by performing a GPA, restoring scale and dividing the restored coordinates by the superimposed coordinates to get an approximation of centroid size. I imagine that there must be a less cumbersome way to do this. *In step 3 above, the restore command in Morpheus, restores the original data and reflected relabelled counterpart. Is there a way that I can also restore the mean configuration using the computed values? I would be also interested in hearing advice on how to deal with the problem deformed or poorly reconstructed specimens. Kind regards, Ciarán ____________________________ Ciarán Brewster, MSc Department of Archaeology University College Cork Western Road Cork Co. Cork, Ireland. email: adhominin[at]me.com web: http://adhominin.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
