My take is that the reason for showing univariate data has to do with what constitutes a sufficient sample size required for generating good multivariate analysis. I believe 68 subjects was probably not enough to generate meaningful MV numbers. And 140 might not be a sufficient sample to create MV analysis that would impress BP.

As I understand it, creating an MV algorithm that describes an insufficient sample size (68 in the case of lympro) may describe that small sample of people but it would likely fall apart when applied to a larger sample. I posted something about this here before -- if you search online regarding what constitutes a sufficient sample size for multivariate analysis the pros can't seem to agree on what that number should be but it increases according to the number of variables in the equation.

- just my opinion fwiw, (I'm not a math guy)