It has been a while since the last update. I have been busy with the first year of my PhD, consulting work, and also have been working on a new web-framework for R, which is called OpenCPU.
OpenCPU provices a free and open platform for statistical computing in the cloud. It is meant as an open, social analysis environment where people can share and run R functions and objects. For more details, visit the websit: www.opencpu.org
A new version of the ggplot web application has been released and is available at the usual servers. New features include advanced data import, integration with Google Documents, a new layout and a lot of new geoms. Furthermore it should be easier to use and load faster. To get started with the new version, visit the info page or watch the demo video.
The technology and development of the ggplot2 web application will also be discussed during a session about R and the web at useR 2010 this summer (abstract).
Update: Added a new use case video dedicated to visualizing time series.
The lme4 and ggplot2 web applications are currently used on a regular basis. Some users just want to give it a try, others manage to create very nice models and plots.
If you used one of the web applications, and you created something interesting, please consider sharing your story. Send me an email with the datafile, a PDF of your result, and maybe a reference or a few lines about what the model/plot represents and how it was helpful. If it is interesting enough, I will make it into a demo-video, and add the example to my website to inspire other users.
In January I will give a talk about Web Development with R for the use-R groups in both San Fransisco (Jan 12) and Los Angeles (Jan 20). You can register (for free) on the meetup website.
Update: Added San Francisco slides and pictures and a video recording of the Los Angeles talk.
After some testing, I am confident to post the latest web application on my site. yeroon.net/ggplot2 is a web application for exploratory graphical analysis. It implements a (big) subset of the plots that can be made with the CRAN package ggplot2... (read more)
The code and data that I demonstrated during the Los Angeles area R users group meeting of December 3, is available here. For those that were not there: The code shows 4 examples of random effects models that can be fit using the R package lme4. The examples cover basic 2 level regression, Repeated Measures, Item Response Theory, and finally a 4-level fully-crossed binomial model.
You guys make me proud! I am so happy to see that from the people who tried my web applications last week, 87.21% is using a quality browser. That is very good news for a web developer. And thank you for visiting!
The code that was used for the minicourse in multilevel modeling using lme4 is available here. I included some additional lines to show some plots that were not shown in the lecture but can be useful. Furthermore, I also added some lines to fit the Rasch model using the ltm package, to show that estimates are identical.
Tuesday November 24 I will give a short seminar at UCLA about R web development. Details will become available on the UCLA stat seminar website. Update: slides are available here.
yeroon.net/lme4 is a webinterface for the popular CRAN package lme4 by Douglas Bates and Martin Maechler. It should make random effects / multilevel modelling available for applied researchers without the need for any specialized software.
lme4 supports both hierarchical and crossed data and many distribution families, making it also a suitable tool for e.g. IRT and repeated measures data. Read more...
Soon after the blog post on revolution computing about Stockplot, many people seem to have found my website, and my little VPS server started complaining about memory. The next application that will be released, lme4, is many times more demanding than Stockplot, and would probably kill this server at some moment. Therefore I decided to temporarily host a high-computing virtual server on Amazon's EC2 cloud.
So far it seems to perform pretty well and be able to handle all requests. However, unfortunately I am not able to afford a server like this for a long time, so if you are willing to contribute and are able to sponsor or host a mirror, please get in touch.
On a personal note: staring September 16 I will be visting the Statistical Department at UCLA for a few months, to study the work of Jan de Leeuw. I might not be able to answer any email for a few days.
Stockplot is a new R web application that plots live stock prices. It gets is data from Yahoo! Finance, and uses the CRAN package ggplot2 to generate on-demand figures.
The historical data are stored locally in a MySQL database, which is updated once a day from Yahoo servers. However, the latest price of a stock, indicated by the red line, is updated on every plot request. A plot can be requested by double clicking a company, or dragging it into one of the workspaces. Once a plot has been returned, the graphtype and timeframe of the plot can easily be changed by right-clicking the plot and selecting an option from the menu. Try it out yourself!
At July 8-10, useR! 2009 will take place at Rennes, France. The conference website shows an overview of speakers and topics.
During the useR! Focus session Interfaces on Thursday, I will share some of my experiences about developing web applications. The presentation is called 'irttool.com', and part of the presentation will be devoted to this application. You can grab the abstract and slides from the useR! 2009 website, or from here: [abs][slides].
I've added a little Demo Video to the IRTtool website that shows how the web application can be used for simple IRT analyses. You can watch it on www.irttool.com by clicking the Demo icon.
You can also download the abstract about IRT tool that was submitted to the useR 2009 Committee.
This week my very first publication, in cooperation with Stef van Buuren, became available online. The article about Stage Line Diagrams, for which the Puberty Plot webapplication was made, can be downloaded from the Early View section of the Statistics in Medicine website.
Feel free to play around with the web application. It's public and does not require installation of any software.