As prolific as the CRAN website is of packages, there are several packages to R that succeeds in standing out for their wide spread use (and quality), Hadley Wickhams ggplot2 and plyr are two such packages.

And today (through twitter) Hadley has updates the rest of us with the news:

just released new versions of plyr and ggplot2. source versions available on cran, compiled will follow soon #rstats

Going to the CRAN website shows that plyr has gone through the most major update, with the last update (before the current one) taking place on 2009-06-23. And now, over a year later, we are presented with plyr version 1, which includes New functions, New features some Bug fixes and a much anticipated Speed improvements.
ggplot2, has made a tiny leap from version 0.8.7 to 0.8.8, and was previously last updated on 2010-03-03.

Me, and I am sure many R users are very thankful for the amazing work that Hadley Wickham is doing (both on his code, and with helping other useRs on the help lists). So Hadley, thank you!

In hierarchical cluster analysis dendrogram graphs are used to visualize how clusters are formed. I propose an alternative graph named “clustergram” to examine how cluster members are assigned to clusters as the number of clusters increases.
This graph is useful in exploratory analysis for non-hierarchical clustering algorithms like k-means and for hierarchical cluster algorithms when the number of observations is large enough to make dendrograms impractical.

A similar article was later written and was (maybe) published in “computational statistics”.

Both articles gives some nice background to known methods like k-means and methods for hierarchical clustering, and then goes on to present examples of using these methods (with the Clustergarm) to analyse some datasets.

Personally, I understand the clustergram to be a type of parallel coordinates plot where each observation is given a vector. The vector contains the observation’s location according to how many clusters the dataset was split into. The scale of the vector is the scale of the first principal component of the data.

Clustergram in R (a basic function)

After finding out about this method of visualization, I was hunted by the curiosity to play with it a bit. Therefore, and since I didn’t find any implementation of the graph in R, I went about writing the code to implement it.

The code only works for kmeans, but it shows how such a plot can be produced, and could be later modified so to offer methods that will connect with different clustering algorithms.

How does the function work: The function I present here gets a data.frame/matrix with a row for each observation, and the variable dimensions present in the columns.
The function assumes the data is scaled.
The function then goes about calculating the cluster centers for our data, for varying number of clusters.
For each cluster iteration, the cluster centers are multiplied by the first loading of the principal components of the original data. Thus offering a weighted mean of the each cluster center dimensions that might give a decent representation of that cluster (this method has the known limitations of using the first component of a PCA for dimensionality reduction, but I won’t go into that in this post).
Finally all of our data points are ordered according to their respective cluster first component, and plotted against the number of clusters (thus creating the clustergram).

My thank goes to Hadley Wickham for offering some good tips on how to prepare the graph.

source("https://www.r-statistics.com/wp-content/uploads/2012/01/source_https.r.txt")# Making sure we can source code from github
source_https("https://raw.github.com/talgalili/R-code-snippets/master/clustergram.r")data(iris)set.seed(250)par(cex.lab=1.5, cex.main=1.2)
Data <-scale(iris[,-5])# notice I am scaling the vectors)
clustergram(Data, k.range=2:8, line.width=0.004)# notice how I am using line.width. Play with it on your problem, according to the scale of Y.

Here is the output:

Looking at the image we can notice a few interesting things. We notice that one of the clusters formed (the lower one) stays as is no matter how many clusters we are allowing (except for one observation that goes way and then beck).
We can also see that the second split is a solid one (in the sense that it splits the first cluster into two clusters which are not “close” to each other, and that about half the observations goes to each of the new clusters).
And then notice how moving to 5 clusters makes almost no difference.
Lastly, notice how when going for 8 clusters, we are practically left with 4 clusters (remember – this is according the mean of cluster centers by the loading of the first component of the PCA on the data)

If I where to take something from this graph, I would say I have a strong tendency to use 3-4 clusters on this data.

But wait, did our clustering algorithm do a stable job?
Let’s try running the algorithm 6 more times (each run will have a different starting point for the clusters)

source("https://www.r-statistics.com/wp-content/uploads/2012/01/source_https.r.txt")# Making sure we can source code from github
source_https("https://raw.github.com/talgalili/R-code-snippets/master/clustergram.r")set.seed(500)
Data <-scale(iris[,-5])# notice I am scaling the vectors)par(cex.lab=1.2, cex.main= .7)par(mfrow =c(3,2))for(i in1:6) clustergram(Data, k.range=2:8 , line.width= .004, add.center.points=T)

Resulting with: (press the image to enlarge it)

Repeating the analysis offers even more insights.
First, it would appear that until 3 clusters, the algorithm gives rather stable results.
From 4 onwards we get various outcomes at each iteration.
At some of the cases, we got 3 clusters when we asked for 4 or even 5 clusters.

Reviewing the new plots, I would prefer to go with the 3 clusters option. Noting how the two “upper” clusters might have similar properties while the lower cluster is quite distinct from the other two.

By the way, the Iris data set is composed of three types of flowers. I imagine the kmeans had done a decent job in distinguishing the three.

Limitation of the method (and a possible way to overcome it?!)

It is worth noting that the current way the algorithm is built has a fundamental limitation: The plot is good for detecting a situation where there are several clusters but each of them is clearly “bigger” then the one before it (on the first principal component of the data).

For example, let’s create a dataset with 3 clusters, each one is taken from a normal distribution with a higher mean:

source("https://www.r-statistics.com/wp-content/uploads/2012/01/source_https.r.txt")# Making sure we can source code from github
source_https("https://raw.github.com/talgalili/R-code-snippets/master/clustergram.r")set.seed(250)
Data <-rbind(cbind(rnorm(100,0, sd=0.3),rnorm(100,0, sd=0.3),rnorm(100,0, sd=0.3)),
cbind(rnorm(100,1, sd=0.3),rnorm(100,1, sd=0.3),rnorm(100,1, sd=0.3)),
cbind(rnorm(100,2, sd=0.3),rnorm(100,2, sd=0.3),rnorm(100,2, sd=0.3)))
clustergram(Data, k.range=2:5 , line.width= .004, add.center.points=T)

The resulting plot for this is the following:

The image shows a clear distinction between three ranks of clusters. There is no doubt (for me) from looking at this image, that three clusters would be the correct number of clusters.

But what if the clusters where different but didn’t have an ordering to them?
For example, look at the following 4 dimensional data:

source("https://www.r-statistics.com/wp-content/uploads/2012/01/source_https.r.txt")# Making sure we can source code from github
source_https("https://raw.github.com/talgalili/R-code-snippets/master/clustergram.r")set.seed(250)
Data <-rbind(cbind(rnorm(100,1, sd=0.3),rnorm(100,0, sd=0.3),rnorm(100,0, sd=0.3),rnorm(100,0, sd=0.3)),
cbind(rnorm(100,0, sd=0.3),rnorm(100,1, sd=0.3),rnorm(100,0, sd=0.3),rnorm(100,0, sd=0.3)),
cbind(rnorm(100,0, sd=0.3),rnorm(100,1, sd=0.3),rnorm(100,1, sd=0.3),rnorm(100,0, sd=0.3)),
cbind(rnorm(100,0, sd=0.3),rnorm(100,0, sd=0.3),rnorm(100,0, sd=0.3),rnorm(100,1, sd=0.3)))
clustergram(Data, k.range=2:8 , line.width= .004, add.center.points=T)

In this situation, it is not clear from the location of the clusters on the Y axis that we are dealing with 4 clusters.
But what is interesting, is that through the growing number of clusters, we can notice that there are 4 “strands” of data points moving more or less together (until we reached 4 clusters, at which point the clusters started breaking up).
Another hope for handling this might be using the color of the lines in some way, but I haven’t yet figured out how.

Clustergram with ggplot2

Hadley Wickham has kindly played with recreating the clustergram using the ggplot2 engine. You can see the result here: http://gist.github.com/439761
And this is what he wrote about it in the comments:

I’ve broken it down into three components:
* run the clustering algorithm and get predictions (many_kmeans and all_hclust)
* produce the data for the clustergram (clustergram)
* plot it (plot.clustergram)
I don’t think I have the logic behind the y-position adjustment quite right though.

Conclusions (some rules of thumb and questions for the future)

In a first look, it would appear that the clustergram can be of use. I can imagine using this graph to quickly run various clustering algorithms and then compare them to each other and review their stability (In the way I just demonstrated in the example above).

The three rules of thumb I have noticed by now are:

Look at the location of the cluster points on the Y axis. See when they remain stable, when they start flying around, and what happens to them in higher number of clusters (do they re-group together)

Observe the strands of the datapoints. Even if the clusters centers are not ordered, the lines for each item might (needs more research and thinking) tend to move together – hinting at the real number of clusters

Run the plot multiple times to observe the stability of the cluster formation (and location)

Yet there is more work to be done and questions to seek answers to:

The code needs to be extended to offer methods to various clustering algorithms.

How can the colors of the lines be used better?

How can this be done using other graphical engines (ggplot2/lattice?) – (Update: look at Hadley’s reply in the comments)

What to do in case the first principal component doesn’t capture enough of the data? (maybe plot this graph to all the relevant components. but then – how do you make conclusions of it?)

What other uses/conclusions can be made based on this graph?

I am looking forward to reading your input/ideas in the comments (or in reply posts).

For your convenience (and with Ian’s permission), I am reposting his proposal here. You are welcome to send him feedback by e-mailing him (at: [email protected]), or by leaving a comment here (and I will direct him to your comment).

I was delighted to see the following e-mail post from Dirk Eddelbuettel regarding the google-summer-of-code R google group:
* * *

Earlier today Google finalised student / mentor pairings and allocations for
the Google Summer of Code 2010 (GSoC 2010). The R Project is happy to
announce that the following students have been accepted:

Colin Rundel, “rgeos – an R wrapper for GEOS”, mentored by Roger Bivand of
the Norges Handelshoyskole, Norway

Ian Fellows, “A GUI for Graphics using ggplot2 and Deducer”, mentored by
Hadley Wickham of Rice University, USA

Chidambaram Annamalai, “rdx – Automatic Differentiation in R”, mentored by
John Nash of University of Ottawa, Canada

Yasuhisa Yoshida, “NoSQL interface for R”, mentored by Dirk Eddelbuettel,
Chicago, USA

Felix Schoenbrodt, “Social Relations Analyses in R”, mentored by Stefan
Schmukle, Universitaet Muenster, Germany

The R Project is honoured to have received its highest number of student
allocations yet, and looks forward to an exciting Summer of Code. Please
join me in welcoming our new students.

At this time, I would also like to thank all the other students who have
applied for working with R in this Summer of Code. With a limited number of
available slots, not all proposals can be accepted — but I hope that those
not lucky enough to have been granted a slot will continue to work with R and
towards making contributions within the R world.

I would also like to express my thanks to all other mentors who provided for
a record number of proposals. Without mentors and their project ideas we
would not have a Summer of Code — so hopefully we will see you again next
year.

Regards,

Dirk (acting as R/GSoC 2010 admin)

* * *

From all the projects, the one I am most excited about is:
Ian Fellows, “A GUI for Graphics using ggplot2 and Deducer”, mentored by Hadley Wickham of Rice University, USA

Deducer (text from the website) attempts to be a free easy to use alternative to proprietary data analysis software such as SPSS, JMP, and Minitab. It has a menu system to do common data manipulation and analysis tasks, and an excel-like spreadsheet in which to view and edit data frames. The goal of the project is to two-fold.

Provide an intuitive interface so that non-technical users can learn and perform analyses without programming getting in their way.

Increase the efficiency of expert R users when performing common tasks by replacing hundreds of keystrokes with a few mouse clicks. Also, as much as possible the GUI should not get in their way if they just want to do some programming.

Deducer is designed to be used with the Java based R console JGR, though it supports a number of other R environments (e.g. Windows RGUI and RTerm).

This combination (of Deducer and ggplot2) might finally provide the bridge to the layman-statistician that some people recently wrote to be one of R’s weak spots (while other bloogers wrote back that this is o.k., still no one refuted that R doesn’t compete with the point-and-click of softwares like SPSS or JMP.)
I came across Ian in the discussion forums, where he provided very kind help to his package “deducer”. Coupled with having Hadley as his mentor, I am very optimistic about the prospects of seeing this project reaching very high standards.
Very exciting development indeed!

yeroon.net/ggplot2 is a web interface for Hadley Wickham’s R package ggplot2. It is used as a tool for rapid prototyping, exploratory graphical analysis and education of statistics and R. The interface is written completely in javascript, therefore there is no need to install anything on the client side: a standard browser will do.

The new version has a lot of cool new features, like advanced data import, integration with Google docs, converting variables from numeric to factor to dates and vice versa, and a lot of new geom’s. Some of which you can watch in his new video demo of the application:

One of the exciting new frontiers for R programming is of creating website interfaces to R code. At the forefront of this domain is a young and (very) bright man called Jeroen Ooms, whom I had the pleasure of meeting at useR 2009 (press the link to see his presentation).

Today Jeroen announced a new version (0.11) of his web interface to ggplot2. See it here: http://www.yeroon.net/ggplot2/

As Jeroen wrote:

New features include 1D geom’s (histogram, density, freqpoly), syntax mode (by clicking the tiny arrow at the bottom), and some additional facet options. And some minor improvements and fixes, most notably for Internet Explorer.
The data upload has not been improved yet, I am working on that. For now, it supports .csv, .sav (spss), and tab delimited data. Please make sure your filename has the appropriate extension and every column has a header in your data. If you export a dataframe from R, use:
write.csv(mydf, ”mydf.csv” , row.names=F). If you upload an spss
datafile, none of this should be a concern.
Supported browsers are IE6-8, FF, Safari, and Chrome, but a recent browser is highly recommended. As always, feedback is more than welcome.

Here is a little demo video that shows how to use the new features: