"All models are wrong, but some are useful" begins the Wired Magazine article The End of Theory.
The article, by Chris Anderson (of The Long Tail fame), goes on to explain that in this current 'era of massively abundant data' scientists have become less reliant on the age old tradition of creating plausible models and testing hypotheses to refute or deny the model.
Instead, scientists can merely find correlations is massive data sets that have been, heretofore, unavailable for study and searching.
The power of sufficient data allows us to say "Correlation is enough" (Anderson).
Physical therapy has embraced this new thinking since 2002.
A manipulation decision rule was published in Spine magazine that laid out 5 predictor variables PT clinicians could test for: the presence of 3/5 variables implied that the patient was a good candidate for manipulation.
Anderson states that massively abundant data
"forces us to view data mathematically first and establish a context for it later."The researchers first measured their physical therapy patients with dozens of 'traditional' physical therapy test and measures.
They then dumped the results of these measurements (some good, some bad) into a statistical 'hopper' and compared the results with the patients that got better.
Those tests that best predicted the patients that got better were the final predictor variables.
To paraphrase Anderson...
Who knows why patients get better? The point is that patients do get better and we can track and measure those patients with increasing accuracy.
"With enough data, the numbers speak for themselves"
Measurements are the key, however. Without measurements we cannot draw correlations.
Even Google could not search web pages if no one bothered to post their data.