Dear Max, Thank you very much ! Your sample code is very helpful.
In linear separable problem, I should use fda by linear regression instead of bruto unless taking some dimensional reduction process, should I? Cheers. Shuji On 2007/04/09, at 22:25, Kuhn, Max wrote: > Shuji, > > I suspect that bruto blows up because your data are linearly > separable. > To see this (if you didn't already know), try > > library(lattice) > splom(~x, groups = y) > > and look at the first row. If you are trying to do classification, > there > are a few methods that would choke on this (logistic regression) and a > few that won't (trees, svms etc). I would guess that bruto is in the > latter group. > > However, if you are try to do classification, try using bruto via fda: > >> tmp <- cbind(x, factor(y)) >> >> fdaFit <- fda(y2~., tmp) >> fdaFit > Call: > fda(formula = y2 ~ ., data = tmp) > > Dimension: 1 > > Percent Between-Group Variance Explained: > v1 > 100 > > Degrees of Freedom (per dimension): 5 > > Training Misclassification Error: 0 ( N = 20 ) >> >> predict(fdaFit, type = "posterior")[1:3,] > 0 1 > 2 0 1 > 2 0 1 > 2 0 1 > > Max > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of ?? ?? > Sent: Sunday, April 08, 2007 10:47 PM > To: [email protected] > Subject: [R] Could not fit correct values in discriminant analysis by > bruto. > > Dear R-users, > > I would like to use "bruto" function in mda package for flexible > discriminant analysis. > Then, I tried, for example, following approach. > >> x > band1 band2 band3 > 1 -1.206780 -1.448007 -1.084431 > 2 -0.294938 -0.113222 -0.888895 > 3 -0.267303 -0.241567 -1.040979 > 4 -1.206780 -1.448007 -1.040979 > 5 -1.151518 -0.806286 -0.671630 > 6 -1.179146 -1.396670 -1.453775 > 7 -0.294938 -0.241567 -1.453775 > 8 -0.350200 -0.267239 -1.084431 > 9 -1.151518 -0.857623 -0.649901 > 10 1.362954 -1.396670 -2.235926 > 11 -0.239675 1.118883 1.457551 > 12 -0.294938 -1.268325 -0.497817 > 13 -0.294938 -0.729278 -0.106745 > 14 -1.123883 -0.703612 -0.150196 > 15 0.616905 1.144548 -0.150196 > 16 -0.267303 1.657930 1.044750 > 17 1.611637 1.041874 0.610225 > 18 -1.123883 -0.677941 0.262605 > 19 -0.239675 -0.626604 -0.128473 > 20 2.274797 1.118883 1.805171 > >> y > [1] 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 > >> fit <- bruto(x,y) > > But, obtained fit$fitted.values are enormously high (or low) . > Execution of bruto(x[,2:3], y) is done well (values are nearly 1 or > 0). > Values of column 1 are wrong or appropriate option is needed? > I contacted the package maintainer, but the problem could not be > solved. > > Thanks > > Shuji Kawaguchi > >> R.version > platform i386-apple-darwin8.8.1 > arch i386 > os darwin8.8.1 > system i386, darwin8.8.1 > version.string R version 2.4.0 (2006-10-03) ______________________________________________ [email protected] 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.
