I was reading this article on SEC QB stats:
http://maroonandwhitenation.com/201...ts-determine-secs-best-returning-quarterback/
The author assigns point values to different quarterback metrics to try to determine the best returning SEC QB. Like many of you, I took a statistics course at State, but I'm far from an expert. The methodology he used seems terribly flawed, but I was hoping someone with more expertise in mathematics or statistics could break down the problems.
A couple of issues that I noticed:
My memory from stats is that each term (category) should have a weight assigned and be normalized to some perfect value (1 or 100 usually)
Can any math/statistics wizards weigh in?
http://maroonandwhitenation.com/201...ts-determine-secs-best-returning-quarterback/
The author assigns point values to different quarterback metrics to try to determine the best returning SEC QB. Like many of you, I took a statistics course at State, but I'm far from an expert. The methodology he used seems terribly flawed, but I was hoping someone with more expertise in mathematics or statistics could break down the problems.
A couple of issues that I noticed:
- He seems to double-count a lot of the measurements. For instance, he assigns a value for rushing yards, passing yards, and total yards. Since total yards is just the sum of rushing and passing, you don't need a separate value for that measure, right?
- There is no normalization or weighting to any of the measurements. So every "sub category" just ends up with the same weight. Categories with more sub categories will be weighted more than other, though I don't think that's what he intended.
My memory from stats is that each term (category) should have a weight assigned and be normalized to some perfect value (1 or 100 usually)
Can any math/statistics wizards weigh in?
