Visualize the difference between PCA and LDA on the iris dataset.

lda_vs_pca.R

require(MASS)
require(ggplot2)
require(scales)
require(gridExtra)

pca <- prcomp(iris[,-5],
              center = TRUE,
              scale. = TRUE) 

prop.pca = pca$sdev^2/sum(pca$sdev^2)

lda <- lda(Species ~ ., 
           iris, 
           prior = c(1,1,1)/3)

prop.lda = r$svd^2/sum(r$svd^2)

plda <- predict(object = lda,
                newdata = iris)

dataset = data.frame(species = iris[,"Species"],
                     pca = pca$x, lda = plda$x)

p1 <- ggplot(dataset) + geom_point(aes(lda.LD1, lda.LD2, colour = species, shape = species), size = 2.5) + 
  labs(x = paste("LD1 (", percent(prop.lda[1]), ")", sep=""),
       y = paste("LD2 (", percent(prop.lda[2]), ")", sep=""))

p2 <- ggplot(dataset) + geom_point(aes(pca.PC1, pca.PC2, colour = species, shape = species), size = 2.5) +
  labs(x = paste("PC1 (", percent(prop.pca[1]), ")", sep=""),
       y = paste("PC2 (", percent(prop.pca[2]), ")", sep=""))

grid.arrange(p1, p2)

Great example! This helped me wrap my head around the lda function. While running this code I noticed a small typo.

16 prop.lda = lda$svd^2/sum(lda$svd^2) <- prop.lda = r$svd^2/sum(r$svd^2)

Great example! Thanks Keele for the correction on line 16
prop.lda = lda$svd^2/sum(lda$svd^2)

thanks for you example!
I have a question

this example compare PCA and LDA
if i want to compare other type of PCA like kernel PCA or PGA with PCA and see its result on LDA,
what can i do?
i mean comparing pca and pga and see what happen in lda result ???

Hi,
I have a question.

My data.frame have a colum with species names (the first colum) and I want to insert these names in each point. Is it possible?

yes, you can simply add "labels = yourcolumname"
at least this will work for ggbiplot()

there is some fault here in line 12

it should be 'r' in place of 'lda' on left side

> lda <- lda(Species ~ ., 
>            iris, 
>            prior = c(1,1,1)/3)
> 

should be like

> 
> r <- lda(Species ~ ., 
>            iris, 
>            prior = c(1,1,1)/3)
> 

Dear @thigm85 ,
I'm getting this error (Error in UseMethod("predict") :
no applicable method for 'predict' applied to an object of class "function") when running

plda <- predict(object = lda,
                newdata = iris)

Any idea how to fix this?

Works fine for me.
Did you load the package correctly?

Dear @thigm85 ,
I'm getting this error (Error in UseMethod("predict") :
no applicable method for 'predict' applied to an object of class "function") when running

plda <- predict(object = lda,
                newdata = iris)

Any idea how to fix this?

Hi, it's a cool code to visualize the PCA and LDA.
But I wonder whether you also have an approach to add the arrows of different discriminators onto the LDA plot, so that the contribution and relationship of these discriminators can be partly read simultaneously.
KR - Qing