Hello all, This is the problem I am dealing with: I have an R data frame with the sample of observations of the control variables "x1", "x2" , ... and the response variable "y". I also have a set of "base" functions = "x1, "x2", ..., "log(x1)", "log(x2)", ... What I need is to have the ANALYTICAL function, i.e. a polynom of the given base functions, (s.t like x1^3*log(x2) ) that represents my data, i.e. that minimizes the average errors of the predictions. I guess that using non-parametric regression I can obtain a non analytical answer. Do you know of any solutions for that problem or similars? If so, where can I find them? Many, many thanks, Casiano
Have you considered stepwise regression, e.g., stepAIC from the MASS library or stepAIC.c dowloadable from "prodsyse.com"? The latter is not solidly debugged but has worked with several examples with "lm". If you have the time, you can dig through the code and modify the hierarchy rules to suit your needs. hope this helps. spencer graves Casiano Rodriguez Leon wrote:> Hello all, > > This is the problem I am dealing with: > > I have an R data frame with the sample of observations of the control > variables "x1", "x2" , ... and the response variable "y". > I also have a set of "base" functions = "x1, "x2", ..., "log(x1)", > "log(x2)", ... > > What I need is to have the ANALYTICAL function, i.e. a polynom of the > given base functions, (s.t like x1^3*log(x2) ) that represents my data, > i.e. that minimizes the average errors of the predictions. > > I guess that using non-parametric regression I can obtain a non analytical > answer. Do you know of any solutions for that problem or similars? If so, > where can I find them? > > Many, many thanks, > > Casiano > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > stat.math.ethz.ch/mailman/listinfo/r-help