I agree with most of the suggestions but there is still a role for relative warps in such a study. They give one an overall view of the major variation in the data. The ordination plot often allows one to detect (and thus eliminate) any wild outliers that might be present. It will also give one some feeling for whether the variation within the a priori defined groups might be more or less homogeneous - as assumed by the suggested MANOCVA.
-------------------- F. James Rohlf - Dept. Ecology & Evolution SUNY, Stony Brook, NY 11794-5245 > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of > [EMAIL PROTECTED] > Sent: Friday, January 30, 2004 8:16 AM > To: [EMAIL PROTECTED] > Subject: response to shark question > > > Javier > Your question is straightforward and I can recommend the > approach I would take. With 19 landmarks you should have 34 > PW's. My philosophy says you should ignore the RWs as they > are an "undesigned" morphospace with respect to questions in > which you are interested. I would perform the following > MANCOVA, being careful that the stats program is set to > center polynomials: > > PW1, PW2... PW34 = intercetp + sex + centroid size + sex*size > > If the interaction term is not significant drop it and re-run > the model. Save canonical scores from both the sex and size > main effects. > > Use tpsRegr with your original landmarks loaded as dependent > variables (tpsRegr will turn these into PWs) and a given set > of canonical scores (e.g. for the sex effect) as the > independent variable. Then you get a visualization of a > canonical axis which represents a "designed morphospace". It > is the thing you explicitly want to visualize, not a > principle component. > > As a side note, 19 is a lot of landmarks if there are only 75 > specimens, so you might consider dropping relatively > redundant points and those that have alot of measurement error. > > My $.02 > -Thom > > ________________><()()(>________________ > > Dr. Thomas J. DeWitt, Assistant Professor > Department of Wildlife & Fisheries Sciences > & Program in Bioenvironmental Sciences > Texas A&M University > 2258 TAMU > College Station, TX 77843-2258 > > Tel. (979) 458-1684 (office) > Tel. (979) 845-7522 (lab) > Fax (979) 845-4096 > E-mail [EMAIL PROTECTED] > Web http://wfscnet.tamu.edu/faculty/tdewitt/webpage.htm > TAMU Map to DeWitt lab & office: > http://www.tamu.edu/map/gifs/detail/FGHB.gif > > >>> <[EMAIL PROTECTED]> 01/29/04 09:29AM >>> > I am doing a morphometric study about a shark, but I've got > difficulties with the interpretation of the results. I've > digitalized 75 specimens and marked 19 landmarks. I don't > know whether is better to see the visuallization plot of the > partial or of the relative warps for the interpretation of > the results. > > I've obtained 16 partial warps and 34 relative warps(RW). The > 3 first RW explain only the 57% of the total change, and the > 90% is explained by the 12 first RW. > > Moreover I think that RW 1 and 3 represent a bending due to > the fact that a slender, long fish is difficult to position > completely straight. I don't know if it is possible to > eliminate this "artificial change" from the analysis in order > to quantify only the real change. (In fact, I'm not sure if > it is reasonable to study shape in slender fish from their > digital images or if it would be better to use the > traditional morphometrics and make the measures from each specimen) > > The main objective of my study is to evaluate the ontogenetic > change and the change between sexes (if it exists). In my > preliminary results I've treated all the specimens as a > whole, not making groups by size or sex. It would be better to do so? > > I've read one paper in which partial warp scores and uniform > were regressed on size to test the hypothesis of ontogenetic > allometry, but I don't understand why size must be used for > testing it. I think that the partial warp scores are the > weight matrix, though I'm not sure. In case I was right, I > should enter the weight matrix into the option data of the > tpsRegr program and the centroid size as an independent > variable, or it has no sense ? > > Any advice or comment to the extense number of questions and > doubts will be welcome. > > Thanks in advance for all your help. > > Javier Mariana (mail: [EMAIL PROTECTED]) > > Barcelona (Spain) > > == > Replies will be sent to list. > For more information see > http://life.bio.sunysb.edu/morph/morphmet.h> tml. > == > Replies > will be sent to list. > For more information > see http://life.bio.sunysb.edu/morph/morphmet.html. > == Replies will be sent to list. For more information see http://life.bio.sunysb.edu/morph/morphmet.html.
