Re:Data mining without prejudice
by tel on Monday December 19, 2011 @02:13PM
(#3370717)
I love these techniques but wish they'd market themselves differently. This is hardly able to detect true relationships between variables---the false positive rate will likely be astronomical even with their MCMC p-value table on the website---but it is a more efficient way of exploring a high-dimensional data set than just comparing it blindly.
I think put in the camp with Hadley Wickham's Grand Tour methodologies, this is pretty interesting. Outside of that, it's just a well-marketed method jut waiting for failure.
I think put in the camp with Hadley Wickham's Grand Tour methodologies, this is pretty interesting. Outside of that, it's just a well-marketed method jut waiting for failure.