Dear Jens, If I understood well your problem you have an issue with total length of individuals you measured. I have not much details from your investigations but my question is: did you consider the length before to perfom your analysis? I mean, did you correct the head measurments (I suppose they are distances) by mean total or mean standard length of whole individuals, before perfoming the PCA analysis? This size effect you observed is quite suspiciuos for me. If you want to observe differences in the effects of different diets, it should be appropriate to eliminate each size dependent effects from your data by using for example the Elliott et al., 1985 formula, quite used in "fisheries" works. I hope this could help you Cheers Claudio
Il giorno lunedì 26 ottobre 2015 14:30:24 UTC+1, Jens De meyer ha scritto: > > Dear, > > I have a question about EFDs, and was wondering if you could help me? I > have performed a PCA on standardized EFDs (standardized for first harmonic, > using SHAPE-software). The EFDs represented head shapes of eels which were > given different diets (we compared three different diets). The results of > the PCA indicated that PC1 was related to variation in head width, with > high PC-scores being related to broader heads and vice versa. Next, we > performed a linear regression on these PC-scores and log(length) of the > three different diets and observed a size-effect in all three diets. > Because of this, I wanted to divide the PC-scores by the total length of > the studied eels. Then, we would perform an ANOVA on these length-corrected > PC-scores between the different diets, as this would allow us to find out > whether the mean head width of a certain diet is significantly broader than > another diet. > Therefore, my question to you is: Am I acutally allowed to divide these > PC-scores by total length? > Many thanks in advance! > Kind regards, > Jens De Meyer -- MORPHMET may be accessed via its webpage at http://www.morphometrics.org To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
