iris.Rmd

---
title: "Illustration of PCAviz on Iris data"
date: "`r Sys.Date()`"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{PCAviz Iris demo}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

This vignette illustrates the PCAviz plotting interface on the Iris
data set. Although this data set is small and near ubiquitously used
for simple demonstrations in R, it serves well to illustrate the range
of plotting options in the PCAviz package.

```{r, echo = FALSE, message = FALSE}
knitr::opts_chunk$set(collapse = TRUE,comment = "#",fig.width = 5,
                      fig.height = 4,fig.align = "center",
                      fig.cap = " ",results = "hold")
```

Begin by loading these packages into your R environment. 

```{r, message = FALSE}
library(PCAviz)
library(ggplot2)
library(magrittr)
# library(rsvd)
```

## Load the data and compute the PCs

Load the Iris data. Here, we add an "id" column to the table.

```{r}
data(iris)
iris <- cbind(iris,data.frame(id = 1:150))
iris <- transform(iris,id = as.character(id))
```

Compute principal components using `prcomp` (alternatively,
you may use `princomp` or `rpca` from the `rsvd` package).

```{r}
# out.pca <- princomp(iris[1:4])
# out.pca <- rpca(iris[1:4],k = 4,center = TRUE,scale = FALSE,retx = TRUE)
out.pca <- prcomp(iris[1:4])
```

Create the "pcaviz" object from the Iris data and the computed PCs.

```{r}
iris <- pcaviz(out.pca,dat = iris)
```

The first PC captures over 90% of the total variance in the Iris
samples.

```{r}
screeplot(iris,type = "pve")
```

The first PC captures more variation in Petal Length than other variables.

```{r, fig.height = 4, fig.width = 4}
pcaviz_loadingsplot(iris,pc.dim = "PC1")
```

Scale the PCs, then rotate the first two PCs by 15 degrees.

```{r}
iris <- pcaviz_reduce_whitespace(iris) %>%
        pcaviz_rotate(15)
```
		
Print out a summary of the PCA results and accompanying Iris data.

```{r}
summary(iris)
```

## Create visualizations of the Iris PCA results

Calling `plot` without any additional options shows the samples projected 
onto the first two PCs, labeled by Species.  The large circles represent the 
group summaries  (that is, the median PC projection for each of the three Iris 
species). Also observe that an abbreviated label is automatically created.

```{r}
plot(iris)
```

Setting `draw.points = TRUE` instead uses different colors and shapes to
depict the "Species" assignment.

```{r}
plot(iris,draw.points = TRUE)
```

In the next plot, we use a color gradient to represent a continuous variable
(Petal Length), and the sample ids are plotted instead of points.

```{r}
plot(iris,color = "Petal.Length",label = "id")
```

Show Petal Length as different colors, and Species as different shapes.

```{r}
plot(iris,draw.points = TRUE,color = "Petal.Length",shape = "Species")
```

Plot the first PC against Petal Width. When plotting a PC against a data
column, the linear relationship between the two variables is automatically
summarized with the linear best fit (dashed line), and the confidence 
intervals (dotted lines).

```{r}
plot(iris,coords = c("PC1","Petal.Width"),draw.points = TRUE)
```

The PCAviz plotting interface also allows for quickly plotting
multiple combinations of PCs. The default, when multiple PCs are
selected, is to plot all combinations of the PCs:

```{r, fig.width = 7,fig.height = 6}
plot(iris,coords = paste0("PC",1:4),group = NULL)
```

An alternative to plotting all PC combinations is to plot pairs of
consecutive PCs, which can be conveniently performed by setting
`arrange.coords = "consecutive.pairs"`:

```{r, fig.width = 7, fig.height = 6}
plot(iris,coords = paste0("PC",1:4),arrange.coords = "consecutive.pairs",
     draw.points = TRUE,group = NULL)
```
	 
The PCAviz package has an easy-to-use interface for creating
violin plots to visualize the relationship between PCs and categorical
data variables. In this example, Species is compared against all four PCs.

```{r}
pcaviz_violin(iris)
```

One advantage of ggplot2 graphics is that it easily allows for combining 
plots; however, one has to take some care to ensure that the ggplot layers
are evaluated correctly. Here we give an example of overlaying additional 
data on top of the PCAviz scatterplot (we use this additional layer in this
example to highlight samples with unusually small or unusually large sepals). 
Note that the `inherit.aes = FALSE` option is needed, otherwise the code 
will generate an error.

```{r}
dat <- subset(iris$data,Sepal.Width <= 2 | Sepal.Width >= 4)
dat <- with(dat,data.frame(x = PC1,y = PC2))
plot(iris,draw.points = TRUE,shape = "Species",group = NULL) +
  geom_point(data = dat,aes(x = x,y = y),
             shape = 1,size = 5.5,inherit.aes = FALSE)
```

Create an interactive plot using the `plotly` package, and embed it in a 
separate HTML file. View the interactive plot [here](iris_plotly.html).

```{r}
iris_plotly <- plot(iris,plotly = TRUE,
                    tooltip = c("id","Species","Sepal.Length","Sepal.Width",
                                "Petal.Length","Petal.Width"),
                    plotly.file = "iris_plotly.html")
```

Note that the interactive plot can also be easily embedded within this 
document. In this example we have placed it in a separate webpage because
loading the JavaScript can be slow in some browsers.