R help - Jan 2012 - Non Negative Least Squares Regression with nnls

Hello R experts,

I have two questions related to the nnls library
(http://www.inside-r.org/packages/cran/nnls), and more broadly to linear
regression with positive coefficients. Sample code is below the Qs.

Q1: Regular regression (with lm) gives me the significance of each variable. How
do I get variable significance with nnls? If there's no ready function, any
easy way to manually derive them?
Q2: With regular regression, I can model interactions by doing (a*b). How do I
model interactions with nnls? Any workarounds to manually perform interactions
if no built-in way?

I'm not a statistician; any links to references are highly appreciated as
are any relevant workarounds or alternative approaches.


Code:
library(nnls)

#data
avgprice <-
c(116.0666667,145.1034483,145.75,131.6666667,113.8163265,122.7142857,120.5882353)
stars <- c(2.5,2.5,2.5,2.5,3,3,3)
lookahead <- c(2,30,60,15,2,30,60)

#regular regression
reg1 <- lm(avgprice ~ stars + lookahead + stars*lookahead)
summary(reg1)

#Non-negative
s<-cbind(stars,lookahead)
reg2<-nnls(s, avgprice)
print(reg2)



Thanks and Regards,
Prasad Joglekar

MBA Candidate, Class of 2012 | Tuck School of Business at Dartmouth
1210 Byrne Hall, Hanover NH 03755
http://www.linkedin.com/in/prasadjoglekar


	[[alternative HTML version deleted]]