Hi, by following these steps you will end up with a valid estimate of repeatability. The website references shows how to pool the standard deviations from each sample unit. It refers to days and runs, in your case these just correspond to different sample units.
1. Select k-1 landmarks on a sample unit. 2. Rearrange both the sample unit and the k-1 landmark-configuration in n random positions (i.e. use the same translation and rotation on the sample unit as the landmark configuration). 3. Select the landmark of interest on the n randomly positioned instances of the sample unit. Ideally this would be done without displaying the locations of the k-1 landmarks already selected, as these could influence the choice and create bias. 4. Register the k landmarks using only the k-1 landmarks. That is, register the full k-configuration using only the Procrustes distances between the (k-1)-configurations. The Procrustes distances between the (k-1)-configurations will obviously minimise to zero, because there is no variation in them. The variation in the kth landmark is all due to repeatability on the sample unit you used. This must be repeated for each sample unit. The standard deviations can then be pooled for an overall estimate of repeatability as in: http://www.itl.nist.gov/div898/handbook/mpc/section4/mpc441.htm However, I do not see how you are going to estimate the biological variability of each landmark. How are you going to know if the biological variability is greater than the repeatability? If it is not, then that landmark is of no value. Semi-landmarks are arbitrary in their initial positions along the outlines so I cannot see how it is possible to calculate repeatability. Regards, Ben -----Original Message----- From: morphmet To: morphmet Sent: 20/07/2005 12:50 Subject: Re: Repeatability Lets see, take a x number of individuals, make three repeated measurements, preferably by different people, and estimate the proportion of the variation that is explained by individual versus all the available variation. Either landmark by landmakr, of all at once in a MANOVA. For semi landmarks, no idea, never worked with those. Good luck. Kim morphmet wrote: >I was simply wondering what would be the best way, the most appropriate >statistical test, to evaluate the repeatability of your landmark placement. > >Furthermore, would this method also work with semi-landmarks? > > >Thanks > >Martin -- Replies will be sent to the list. For more information visit http://www.morphometrics.org _________________________________________________________________ MSN Search Ireland has launched - test-drive it today! http://search.msn.ie -- Replies will be sent to the list. For more information visit http://www.morphometrics.org
