Post date: Feb 16, 2018 1:42:24 PM
Pedro working with Valentina and Carl has access to imagery from a longitudinal study of the brains of European teenagers, a minority of whom are marijuana users. They want to measure how brain structure correlates with marijuana use. The dataset Pedro will be working with was "cleaned" by omitting blurry images, which is caused by the subject moving while the brain is being imaged.
But this raises a question. Are the omitted images different from those that were retained? If too much THC causes someone to be twitchy and more likely to produce a blurry image—or if the THC users are calmer and it is the nonusers who are more twitchy—are we skewing the results by omitting the images that were blurry?
In this case the data is the imagery, and it is bad because it is blurry. We want to give a practical answer to the question: how can Pedro make quantitative measurements of brain structure from images that are blurry?