----- Forwarded message from Matthew Smith ----- Date: Wed, 27 Jun 2012 16:16:07 -0400 From: Matthew Smith Reply-To: Matthew Smith Subject: Disparity and Procrustes Distance To: morphmet@morphometrics.org
I have managed to confuse myself past the point of being able to navigate my own way back to clarity. I am attempting to calculate both Procrustes distance and Disparity in excel for a few of my data sets. They are set up as follows: Procrustes Distance: Ind TRT Day AlignedX1 AlignedY1 Aligned X2 AlignedY2, etc......... I have 78 individuals measured over several time periods. I want to calculate the distance between shape configurations for each individual from one time to another. I have the matrices matched up and have subtracted one set of aligned coordinates from another. The new 'difference' matrix is then used to determine Procrustes Distance. In excel for each row I have the formula =sqrt(mmult(transpose(a)a)), where a is the difference between vectors. For Disparity: I have the data arranged in the same manner (individuals as rows with their aligned coordinates). I have group means (TRT by age class) and subtracted each individual's aligned coordinates from their group's mean. The new 'centered' matrix of aligned coordinates now represents the distance of each coordinate from their group's mean. Here is where my confusion lies, in theory disparity should be the sum of the Procrustes residuals for all individuals in a group. For the Procrustes residuals I an excel formula of mmult(a,transpose(a)). Those values are summed for each group and divided by n-1. F ratios are calculated and p values are determined through permutation procedures. There is a discrepancy between what I have been reading and the old excel template files I had set up for these methods. I wanted to double check my matrix formulas for excel before analyzing these data since matrix/vector multiplication is no cumulative and if I put the transpose in the wrong spot, it will be incorrect. Any insight here would be greatly appreciated. I know there are other packages out there for this, but for now I prefer to keep working in excel. Thank you in advance. Matthew Smith PhD Candidate Department of Biological Sciences 1 University of Arkansas Science Engineering Building Fayetteville, AR 72701-1201 479-575-2963 mts...@uark.edu ----- End forwarded message ----- ----- End forwarded message -----
