You can use ^2 to get all 2 way interactions and ^3 to get all 3 way interactions, e.g.:
lm(Sepal.Width ~ (. - Sepal.Length)^2, data=iris) The lm.fit function is what actually does the fitting, so you could go directly there, but then you lose the benefits of using . and ^. The Matrix package has ways of dealing with sparse matricies, but I don't know if that would help here or not. You could also just create x'x and x'y matricies directly since the variables are 0/1 then use solve. A lot depends on what you are doing and what questions you are trying to answer. -- Gregory (Greg) L. Snow Ph.D. Statistical Data Center Intermountain Healthcare greg.s...@imail.org 801.408.8111 > -----Original Message----- > From: Matthew Douglas [mailto:matt.dougla...@gmail.com] > Sent: Tuesday, March 01, 2011 1:09 PM > To: Greg Snow > Cc: r-help@r-project.org > Subject: Re: [R] Regression with many independent variables > > Hi Greg, > > Thanks for the help, it works perfectly. To answer your question, > there are 339 independent variables but only 10 will be used at one > time . So at any given line of the data set there will be 10 non zero > entries for the independent variables and the rest will be zeros. > > One more question: > > 1. I still want to find a way to look at the interactions of the > independent variables. > > the regression would look like this: > > y = b12*X1X2 + b23*X2X3 +...+ bk-1k*Xk-1Xk > > so I think the regression in R would look like this: > > lm(MARGIN, P235:P236+P236:P237+....,weights = Poss, data = adj0708), > > my problem is that since I have technically 339 independent variables, > when I do this regression I would have 339 Choose 2 = approx 57000 > independent variables (a vast majority will be 0s though) so I dont > want to have to write all of these out. Is there a way to do this > quickly in R? > > Also just a curious question that I cant seem to find to online: > is there a more efficient model other than lm() that is better for > very sparse data sets like mine? > > Thanks, > Matt > > > On Mon, Feb 28, 2011 at 4:30 PM, Greg Snow <greg.s...@imail.org> wrote: > > Don't put the name of the dataset in the formula, use the data > argument to lm to provide that. A single period (".") on the right > hand side of the formula will represent all the columns in the data set > that are not on the left hand side (you can then use "-" to remove any > other columns that you don't want included on the RHS). > > > > For example: > > > >> lm(Sepal.Width ~ . - Sepal.Length, data=iris) > > > > Call: > > lm(formula = Sepal.Width ~ . - Sepal.Length, data = iris) > > > > Coefficients: > > (Intercept) Petal.Length Petal.Width > Speciesversicolor > > 3.0485 0.1547 0.6234 - > 1.7641 > > Speciesvirginica > > -2.1964 > > > > > > But, are you sure that a regression model with 339 predictors will be > meaningful? > > > > -- > > Gregory (Greg) L. Snow Ph.D. > > Statistical Data Center > > Intermountain Healthcare > > greg.s...@imail.org > > 801.408.8111 > > > > > >> -----Original Message----- > >> From: r-help-boun...@r-project.org [mailto:r-help-bounces@r- > >> project.org] On Behalf Of Matthew Douglas > >> Sent: Monday, February 28, 2011 1:32 PM > >> To: r-help@r-project.org > >> Subject: [R] Regression with many independent variables > >> > >> Hi, > >> > >> I am trying use lm() on some data, the code works fine but I would > >> like to use a more efficient way to do this. > >> > >> The data looks like this (the data is very sparse with a few 1s, -1s > >> and the rest 0s): > >> > >> > head(adj0708) > >> MARGIN Poss P235 P247 P703 P218 P430 P489 P83 P307 P337.... > >> 1 64.28571 29 0 0 0 0 0 0 0 0 0 0 > >> 0 0 0 > >> 2 -100.00000 6 0 0 0 0 0 0 0 1 0 0 > >> 0 0 0 > >> 3 100.00000 4 0 0 0 0 0 0 0 1 0 0 > >> 0 0 0 > >> 4 -33.33333 7 0 0 0 0 0 0 0 0 0 0 > >> 0 0 0 > >> 5 200.00000 2 0 0 0 0 0 0 0 0 0 0 > >> -1 0 0 > >> 6 -83.33333 12 0 -1 0 0 0 0 0 0 0 0 > >> 0 0 0 > >> > >> adj0708 is actually a 35657x341 data set. Each column after "Poss" > is > >> an independent variable, the dependent variable is "MARGIN" and it > is > >> weighted by "Poss" > >> > >> > >> The regression is below: > >> fit.adj0708 <- lm( adj0708$MARGIN~adj0708$P235 + adj0708$P247 + > >> adj0708$P703 + adj0708$P430 + adj0708$P489 + adj0708$P218 + > >> adj0708$P605 + adj0708$P337 + .... + > >> adj0708$P510,weights=adj0708$Poss) > >> > >> I have two questions: > >> > >> 1. Is there a way to to condense how I write the independent > variables > >> in the lm(), instead of having such a long line of code (I have 339 > >> independent variables to be exact)? > >> 2. I would like to pair the data to look a regression of the > >> interactions between two independent variables. I think it would > look > >> something like this.... > >> fit.adj0708 <- lm( adj0708$MARGIN~adj0708$P235:adj0708$P247 + > >> adj0708$P703:adj0708$P430 + adj0708$P489:adj0708$P218 + > >> adj0708$P605:adj0708$P337 + ....,weights=adj0708$Poss) > >> but there will be 339 Choose 2 combinations, so a lot of independent > >> variables! Is there a more efficient way of writing this code. Is > >> there a way I can do this? > >> > >> Thanks, > >> Matt > >> > >> ______________________________________________ > >> R-help@r-project.org mailing list > >> https://stat.ethz.ch/mailman/listinfo/r-help > >> PLEASE do read the posting guide http://www.R-project.org/posting- > >> guide.html > >> and provide commented, minimal, self-contained, reproducible code. > > ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.