## On ldr R Package

If you are running R, connected to the internet, and have write privileges on your computer's disk, you can install ldr by typing the following command in the R command window:

``` > install.packages("ldr") ```

To use the ldr package you must enter the library command in R:

``` > library(ldr) ```

### Some Examples Using ldr

For more detail, see the manuscript about the ldr package.

#### Example 1: Covariance Reduction - CORE

``````# obtain the sample covariances
Sigmas <- list()

#Fit core through ldr using a simulated annealing procedure with GrassmannOptim; short = FALSE
fit <- ldr(Sigmas=Sigmas, ns=c(139, 90), numdir=4, model="core", short=FALSE, sim_anneal=TRUE, max_iter=500)

#Fit core through ldr using a simulated annealing procedure with GrassmannOptim; short = TRUE
fit1 <- ldr(Sigmas=Sigmas, ns=c(139, 90), numdir=2, model="core", verbose=TRUE, short=TRUE, sim_anneal=TRUE, max_iter=500)
summary(fit1)

``````

#### Example 2: Likelihood Acquired Directions - LAD

``````data(flea)
X <- flea[,-1]
y <- as.integer(factor(flea[,1], levels=unique(flea[,1])))
fit <- ldr(X=as.matrix(X), y=y, numdir=3, model="lad", short=FALSE)

mycol <- y
mycol[mycol==1]<-"black"
mycol[mycol==2]<-"blue"
mycol[mycol==3]<-"green"
R <- t(fit\$Gammahat[[3]])%*%t(X)

# Plotting the first two directions of the reductions

#-------------

data(bigmac)
fit <- ldr(X=bigmac[,-1], y=log(bigmac[,1]), ycat=FALSE, numdir=1, nslices=2, model="lad", short=TRUE, verbose=TRUE, sim_anneal=TRUE, max_iter=300)
summary(fit)

# Plotting the response against the reduction
plot(y, t(fit\$Gammahat)%*%t(X), ylab="Sufficient Reduction", xlab="bigmac", cex=1.5)

``````

#### Example 3: Principal Fitted Components - PFC

``````data(bigmac)
# Fit PFC with a piecewise constant basis with 5 slices and short=FALSE
fit <- ldr(X=bigmac[,-1], fy=bf(bigmac[,1], case="pdisc", degree=0, nslices=5), model="pfc", numdir=3, structure="aniso", short=FALSE)
summary(fit)

# Similar fit with short=TRUE
fit1 <- ldr(X=bigmac[,-1], fy=bf(bigmac[,1], case="pdisc", degree=0, nslices=5), model="pfc", numdir=2, structure="aniso", short=TRUE)
summary(fit1)

# Plotting the first direction of the reduction against the response
R <- t(fit1\$Gammahat)%*%solve(fit1\$Deltahat)%*%t(X)
plot(y, R[1,], ylab="Dir1 - PFC", xlab="bigmac", cex=1.5)

``````

The views and opinions expressed in this page are strictly those of the page author. The contents of this page have not been approved by the University of Maryland.