*crossposting from drugmonkey.wordpress.com.*

Yay!

He has a new post up in response to a request for the full regression analysis. This analysis is tasty and here was the bit that drew my attention:

A principal component analysis reveals that a single principal component accounts for 71% of the variance in the overall impact scores. This principal component includes substantial contributions from all five criterion scores, with weights of 0.57 for approach, 0.48 for innovation, 0.44 for significance, 0.36 for investigator and 0.35 for environment.

Is Dr. Berg under the impression that these variables are orthogonal? The correlation, translated into plain English, means that when the reviewers perceive the environment as less than stellar, they may tend to think that the approach must be shitty too. Check the orthogonality, Dr. Berg!