For a school project, I am trying to determine similarity between participants based on different aspects of their athletic mechanics. I have the same number of observations per participant, and 8 variables per observation.
I need to algorithmically determine weights for each variable. My current idea is to weight each variable by the variance of the variable across all observations. My reasoning is that variables that have greater variability will be better indicators of the similarity of participants, since variables with low variance will just clump all of the participants together. I've set the weight of the most varied variable to 1, and all others are weighted relative to this variable.
Would this be considered a reasonable way to determine weights? Is there a peer-reviewed method of weighting similar types of variables? I could not find one that I clearly understood.
Thanks in advance!