Tuesday 01 December 2009 at 5:25 pm
1. Download the package
mysql-workbench-oss-5.1.16-16-1-i386.deb, or other version from
http://dev.mysql.com/downloads/workbench/5.1.html
2. Install some necessary packages with
sudo apt-get install liblua5.1-0 libzib1
3. Install MySQL Workbench with
dpkg -i mysql-workbench-oss-5.1.16-16-1-i386.deb
4. Because of the
bug create a symbolic link with
sudo ln -s /usr/lib/libmysqlclient_r.so.16.0.0 /usr/lib/libmysqlclient_r.so
or
sudo ln -s /usr/lib/libmysqlclient_r.so.15.0.0 /usr/lib/libmysqlclient_r.so
Tuesday 27 October 2009 at 4:00 pm
Two nice pieces on the concept of explanation in science, and other things, by
David Deutsch:
1.
A new way to explain explanation
2.
on our place in the cosmos
His fantastic book
The Fabric of Reality is also worth recommending.
Wednesday 09 September 2009 at 02:09 am
I found it recently on YouTube, a true gem. Leonard Bernstein rehearses Rite of Spring at Schleswig-Holstein Music Festival in 1987,
here.
Tuesday 10 March 2009 at 4:06 pm
Just came across
this nice compilation. If you move the mouse over the individual elements a pop-up with an example will show up.
Thursday 04 December 2008 at 12:37 am
See
here. The Catholic Church does not change its position regarding evolution.
Some years ago there were news (see
here and
here) about square watermelons. Japanese farmers were concerned with the packing and storage inefficiency of the ball-shaped fruit. So they started to grow them in cubic glass boxes. Square-shaped watermelons are easier to store and transport.
Does the fact that the square watermelons did not evolve naturally be an evidence against the theory of intelligent design? ;)
Tuesday 11 November 2008 at 02:42 am
Finally I've put together a page on my research projects. I will try to update it soon. For now I've put some conference slides and visualizations from the network coordination project (I wrote about it before
here). I also uploaded the draft version of the paper.
Follow the link on the right or go directly
here.
Friday 03 October 2008 at 12:05 pm
Do you usually start R with a desktop icon or some other shortcut? Are you tired of using setwd and getwd each time after you start R to get the working directory correctly? If so, then your days of suffering might be just coming to an end.
Having the working directory set correctly is very convenient. You can both read and write files to the proper place without typing (on Windows, usually very long) path names. There are couple of solutions:
- Use
setwd in scripts.
One way to achieve this is to have a setwd function call at the top of your scripts. You then run it every time you do the computations in that script. For example to have at the top of a file a following line:
setwd("c:/path/to/my/directory/")
It is a nice approach, but things get complicated if you move files to different computers, say from home to your office, and have different directory structures, disk names etc. Of course you can change it every time. Or perhaps keep couple of versions and have all of them but one commented, for example:
# setwd("c:/path/to/my/directory/at/home")
setwd("c:/path/to/my/directory/at/work")
Which is also OK, but for me is too much micromanagement. Also, it becomes a problem if the script is not intended for interactive use.
- Use Windows shortcuts.
An alternative might be work with Windows shortcuts for starting R. In shortcut's properties there is a "Start in" field in which you can put the path to the desired folder. If you start R with the modified icon then R's working directory will be correctly set. With that approach you can have, say, couple of R icons on your desktop, each to different project folders.
This is convenient unless you work on 10 projects. Each time you may have to create yet another shortcut.
- Use PATH environment variable.
Another approach is to set the environment PATH variable. If you add the path to R's executable to it then you will be able to start R from whatever directory in the system you want.
To modify the PATH variable you need to right-click on "My Computer" and select "Properties" then go to "Advanced" tab and "Environment variables" button. The way to modify the PATH variable depends on where you installed R. Usually it is something like c:\program files\R\R 2.7.0\bin.
I use this approach myself with Total Commander and its command line. Wherever I am on the disk I can start R in that directory just bu typing rgui and pressing Enter. You can also use Windows Console (cmd) for that.
- Make a context menu option.
Yet another way is to add a command to your context menu (the one appearing when you right-click on things). By right-clicking on a folder and choosing "R" option you can start R with that folder set as the working directory.
To set up such a command you have to modify Windows Registry and will require, I believe, administrative privileges. Look here for details how to do this.
Any other ideas or suggestions?
Tuesday 23 September 2008 at 09:28 am
I have moved my rgraph6 R package to R-Forge.
R-Forge is a website that facilitates development of R packages by providing services for version control (through
Subversion), automatic checking and building of the packages including binaries for Windows and MacOS, as well as for collaboration with other R users/developers.
The
rgraph6 package has been already available through my
private mini-repository. It provides an interface to a pretty compact format for storing undirected graphs as sequences of printable ASCII characters which is quite useful for handling large libraries of undirected graphs. The format itself is due to
Brendan McKay. The detailed description of it is available
here and is also included within the rgraph6 package.
Two crucial functions of the package are written in C. As my knowledge of C is rather low it might be far from perfect. If you know C well you are more than welcome to have a look at the sources and suggest some improvements. I believe that one of the crucial things is checking for the size of the character sequences that are converted to binary numbers and then to decimal. I'm not sure whether it will work for arbitrary network sizes. I plan to put a public advertisement on R-Forge to look for people who would be willing to do that. Actually that was one of the reasons I opened the package development to others through R-Forge. So don't be shy and go ahead! :)
From now on I will not release any future version of rgraph6 through this website. All will be distributed through
rgraph6's website on R-Forge. The older versions will still be available though. You can install the current version from R-Forge directly from R with:
install.packages("rgraph6",repos="http://R-Forge.R-project.org")
Saturday 30 August 2008 at 12:05 am
Earlier in August I was on the R user conference which this year took place in Dortmund. It was a quite an exciting event gathering around 500 people from around the globe and featuring 170 presentations and talks. Topics varied from new developments in the R system, newly implemented statistical methods as well as plethora of very interesting, and often visually nice, applications. You can find the list of abstracts and also slides for majority of the talks on the
conference's website.
I left the conference, I'm sure similarly to others, with a wealthy bag of new R "tricks". The next UseR is planned to take place next summer in Rennes (France). The
website is already there.
Among the available slides are the slides from my talk which was derived from my recent paper on coordination in dynamic social networks in heterogeneous groups. This talk focused on a computer simulation I performed for the paper. You can find the slides
here. Additionally
here (QuickTime, 37Mb) is a movie based on the simulation I describe there. More on the paper itself soon.
Sunday 03 August 2008 at 12:36 pm
The
Wall Street journal reports on the onesty of participants of US polls. Especially in the context of race. Suprisingly, there seems to be no evidence of "interviewer effect": effects of interviewer's characteristics on interviewee's responses. As WSJ reports the race of the interviewer does not have any effect on responses on US presidential voting.
During my studies I remember doing some small analyzes based on
PGSS in which I analyzed some question on gender roles (like whether or not the family role of a women is to guarantee succesfull career for a man). There were huge differences depending on the gender of the interviewer and gender of the interviewee.
Another interesting point in the WJS article is the issue of anonymity. As researchers from University of Colorado claim, without the anonymity the respondendt feel more accountable for their answers and tend to answer more truthfully.
Perhaps also we should stop doing face-to-face interviews as some studies show, that giving answers to a computer provides more onest results...
