tag:blogger.com,1999:blog-38600807.post8978213052640301911..comments2022-10-29T04:20:46.650-04:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Win Probability, In Color!Unknownnoreply@blogger.comBlogger4125tag:blogger.com,1999:blog-38600807.post-87374526987728453932012-10-21T22:39:36.439-04:002012-10-21T22:39:36.439-04:00I made my own loess algorithm as an Excel macro, s...I made my own loess algorithm as an Excel macro, so I don't go to the trouble of doing an r-squared. In R I think it will do it for you. I've been using the np package, which will generate r-squareds. Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-22177442795747451292012-10-21T12:25:56.485-04:002012-10-21T12:25:56.485-04:00I've been using loess, too - I don't suppo...I've been using loess, too - I don't suppose you've seen any way to get an adjusted-R^2? summary gives me an "equivalent number of parameters," so I guess I could tally the residuals myself, but it seems odd to me that I haven't found something like that.Stevenhttps://www.blogger.com/profile/06090531523789747157noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-28036765904206319702012-10-17T12:23:24.811-04:002012-10-17T12:23:24.811-04:00Thanks, Andrew. Right now, I'm relying on the ...Thanks, Andrew. Right now, I'm relying on the default bandwidth functions. The np documentation is sparse and anything beyond LOESS is fairly varsity stuff for me.<br /><br />If there is anyone with expertise in R and the np package and is willing to answer some questions drop me a line!Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-2533757965823481862012-10-17T10:49:39.773-04:002012-10-17T10:49:39.773-04:00You might find "Numerical Methods that Work&q...You might find "Numerical Methods that Work" by Acton interesting.<br /><br />Time and yards come in different units, which is the sort of fact that can become awfully inconvenient when doing kernel methods; the metric is important.Andrew Folandhttp://nuclearmangos.blogspot.comnoreply@blogger.com