# dl.combine: Combine output of several methods In directlabels: Direct Labels for Multicolor Plots

## Description

Apply several Positioning methods to the original data frame.

## Usage

 `1` ```dl.combine(...) ```

## Arguments

 `...` Several Positioning Methods.

## Value

A Positioning Method that returns the combined data frame after applying each specified Positioning Method.

## Author(s)

Toby Dylan Hocking

## Examples

 ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98``` ```## Simple example: label the start and endpoints library(nlme) library(lattice) ratplot <- xyplot( weight~Time|Diet,BodyWeight,groups=Rat,type='l',layout=c(3,1)) both <- dl.combine("first.points","last.points") rat.both <- direct.label(ratplot,"both") print(rat.both) ## same as repeated call to direct.label: rat.repeated <- direct.label(direct.label(ratplot,"last.points"),"first.points") print(rat.repeated) ## same with ggplot2: if(require(ggplot2)){ rp2 <- qplot( Time,weight,data=BodyWeight,geom="line",facets=.~Diet,colour=Rat) print(direct.label(direct.label(rp2,"last.points"),"first.points")) print(direct.label(rp2,"both")) } ## more complex example: first here is a function for computing the ## lasso path. mylars <- function ## Least angle regression algorithm for calculating lasso solutions. (x, ## Matrix of predictor variables. y, ## Vector of responses. epsilon=1e-6 ## If correlation < epsilon, we are done. ){ xscale <- scale(x) # need to work with standardized variables b <- rep(0,ncol(x))# coef vector starts at 0 names(b) <- colnames(x) ycor <- apply(xscale,2,function(xj)sum(xj*y)) j <- which.max(ycor) # variables in active set, starts with most correlated alpha.total <- 0 out <- data.frame() while(1){## lar loop xak <- xscale[,j] # current variables r <- y-xscale%*%b # current residual ## direction of parameter evolution delta <- solve(t(xak)%*%xak)%*%t(xak)%*%r ## Current correlations (actually dot product) intercept <- apply(xscale,2,function(xk)sum(r*xk)) ## current rate of change of correlations z <- xak%*%delta slope <- apply(xscale,2,function(xk)-sum(z*xk)) ## store current values of parameters and correlation out <- rbind(out,data.frame(variable=colnames(x), coef=b, corr=abs(intercept), alpha=alpha.total, arclength=sum(abs(b)), coef.unscaled=b/attr(xscale,"scaled:scale"))) if(sum(abs(intercept)) < epsilon)#corr==0 so we are done return(transform(out,s=arclength/max(arclength))) ## If there are more variables we can enter into the regression, ## then see which one will cross the highest correlation line ## first, and record the alpha value of where the lines cross. d <- data.frame(slope,intercept) d[d\$intercept<0,] <- d[d\$intercept<0,]*-1 d0 <- data.frame(d[j,])# highest correlation line d2 <- data.frame(rbind(d,-d),variable=names(slope))#reflected lines ## Calculation of alpha for where lines cross for each variable d2\$alpha <- (d0\$intercept-d2\$intercept)/(d2\$slope-d0\$slope) subd <- d2[(!d2\$variable%in%colnames(x)[j])&d2\$alpha>epsilon,] subd <- subd[which.min(subd\$alpha),] nextvar <- subd\$variable alpha <- if(nrow(subd))subd\$alpha else 1 ## If one of the coefficients would hit 0 at a smaller alpha ## value, take it out of the regression and continue. hit0 <- xor(b[j]>0,delta>0)&b[j]!=0 alpha0 <- -b[j][hit0]/delta[hit0] takeout <- length(alpha0)&&min(alpha0) < alpha if(takeout){ i <- which.min(alpha0) alpha <- alpha0[i] } b[j] <- b[j]+alpha*delta ## evolve parameters alpha.total <- alpha.total+alpha ## add or remove a variable from the active set j <- if(takeout)j[j!=which(names(i)==colnames(x))] else c(j,which(nextvar==colnames(x))) } } ## Calculate lasso path, plot labels at two points: (1) where the ## variable enters the path, and (2) at the end of the path. if(require(lars)){ data(diabetes,envir=environment()) dres <- with(diabetes,mylars(x,y)) P <- xyplot(coef~arclength,dres,groups=variable,type="l") mylasso <- dl.combine("lasso.labels", "last.qp") plot(direct.label(P,"mylasso")) } ```

directlabels documentation built on Jan. 16, 2021, 5:05 p.m.