I didn't respond to your post because it was very stupid. Many people on here have been pretending they have statistical competence, simply because they took some courses in college. No I didn't invent the term "advanced metrics," I don't know why you are chuckling, pretty much all sports science boutiques use that phrase.
Most all modern statistical analysis in pro sports is done with linear regression, because the theorized relationships between the Y variable (usually wins) and the various X variables are typically modeled as linear relationships. Not curvilinear.
The distributions for most advanced metrics are normal, in response to those who are feigning expertise and suggesting alternative distributions. The only time curvilinear regressions are useful in sports analysis, generally, is when we are analyzing physical movement as a variable, like with UZR in baseball.
You guys are not asking good faith questions. No one who knows what they are talking about would suggest a curvilinear fit for an advanced metric like DVOA. The reason is this: many of the variables in sports are actually discrete and theoretical (an assist for example) as opposed to continuous and physical.
Instead, you guys are simply tossing around whatever phrases you can sort of remember from statistics class. You yourself posted whatever random and half baked concepts you could remember from stat class, garbling most of it as you tried to be clever. Your commentary wasn't a question, it wasn't a statement, it wasn't an argument, it wasn't an idea: you simply posted the statistical material you covered, as an amateur, in college. You cannot even articulate a cogent question about empirical sports analysis. Your question about multivariable regression proves that you don't have a clue. You said:
"I'm not sure what type of analysis fanfanclubclub was doing. He claims he uses 'linear regression'. There are several different types of regression methods, the simplest being linear. I actually like Multivariate Regression," - SeattleCuse
what a joke: all regressions are multivariate.
This is simply beyond clueless, you obviously don't understand that linear regression is multi-variable. Do you think I'm running regressions with a single explanatory variable? Haha, how insanely funny. Obviously, you are confused, and you think that "multivariate" regression is just one of several kinds of regression. You are laughably incorrect.
You are a perfect example of the Luddite posters critiquing me on linear regression at every chance: just faking knowledge so you can hurl blind insults. Why would I answer your other ridiculous questions, when it is obvious that you don't know linear regression at all? Do I really have to tell you what the response variables are? Wins and points.
And you are surprised I didn't respond to your other "super clever" questions? None of you are making serious inquiry into the process of analyzing sports empirically. If you guys remembered what heteroskedasticity was, you'd be asking me all sorts of nonsense questions about heteroskedasticity, simply because you happened to remember the word.
All and any of you who are pretending that linear regression is outdated (I heard someone pretend that LR was a 1970's statistic) are simply clueless. Nearly all modern hypothesis in the hard sciences are analyzed using LR, check with the National Academy of Sciences. It's not like you guys are PhD's asking honest and insightful questions.
