Hi,
I have a measurement of the length of phalanges of gorillas, chimpanzees, and humans, and I want to know whether there are significant differences in this variable among these species, after correction by body size. Thus, I want to use an ANCOVA, where:
Dependent variable: Length of phalanges
Independent variable: species
Covariate: body size.
However, I don't have the body size of each of the individuals I am analyzing, but just an average body size per sex and species (e.g. mean body size for female gorillas is 71 kg). My question is, it's wrong if I use the mean of body size per species and sex as a covariate? Body size is a continuous variable, but it will be the same value for all female gorillas, the same for all male gorillas and so on. It does not vary within these groups, is that a problem?
I'll appreciate any comments!
The mean body size will be perfectly confounded with species:sex. I would guess that you would reduce the possibility to identify differences between the species if both, species:sex and mean body size are in the model. I would fit the model using only species, sex and their interaction. And I would ask it it was possible next time to get individual body size measurements...
Worth also saying what "after correction" means, and if you think this is something Ancova does (see, e.g., http://meehl.umn.edu/sites/meehl.dl.umn.edu/files/084nuisancevariables.pdf). But, as Jochen says, next time it would be good to have the individual measurements. Also, presumably you have some idea how the 3d variable like weight should be transformed to predict a length variable. I don't know about the context so don't know, but linear might not be appropriate.
In the language of multilevel modeling, average body size is a level 2 variable, whereas the other variables are level 1 variables.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Modeling
- Analyses of Variance