Jud,
I used the Golf Digest category scores to determine which of the 7 categories played a bigger or lesser role in a course's total score. The Digest formula gives equal weights to 6 of the categories with x2 for Shot Values, so on average a given category score would be 12.5% of its total score. But courses each have their own characteristics that vary by category (for example, 13.82% of Double Eagle's total score comes from the Conditioning category).
For the similarity rankings, I simply used a given course's relative category weights and applied it to all other courses individual category scores and tracked how each performed relative to the straightforward 12.5% weights. For the most part they make sense, but I'm sure you can isolate examples that on the surface don't compute, like the one you pointed out. In the case of Shadow Creek's, here are its relative category weights:
Shot Values 12.30%
Resistance to Scoring 11.69%
Design Variety 12.70%
Memorability 12.88%
Aesthetics 13.00%
Conditioning 12.36%
Ambience 12.78%
Its relative category weights are similar to National's:
Shot Values 12.26%
Resistance to Scoring 11.57%
Design Variety 12.80%
Memorability 13.01%
Aesthetics 12.88%
Conditioning 12.04%
Ambience 13.15%
In the GD ratings, Resistance to Scoring, Aesthetics and Conditioning are the three that vary most from course to course. So a course that shares the same relative strength or weakness in those categories is likely to show up in these top 10 similarity ranks.
This was merely a mechanical exercise that leverages the category scores to define what makes these top courses great - so therefore it assumes that they are the right categories and amounts, which is a pretty big assumption, especially for a unique course like NGLA.