Monte carlo simulations in R

After a short summer break, the third meeting oft he TRUG took place on October 7th.

Henk Broekhuizen presented how R can be used for Monte Carlo simulations. Henk introduced the problem of combining probability distributions, then introduced the main ideas behind Monte Carlo simulations and finished with some outcomes from his own work in probabilistic MCDA models. In this kind of decision analysis, model outcomes are a complex function of the inputs. When these inputs are probability distributions, calculating the outcomes analytically becomes hard and sometimes impossible. Monte Carlo simulations are a useful and straightforward approach to approximate and visualize these model outcomes.  A pdf file of the presentation can be downloaded here: Third_TRUG_meeting


Resampling methods in R

The second meeting of the TRUG took place on July 11th.

Inga Schwabe gave an introduction into the world of resampling methods. Resampling methods can be a good alternative when classical statistical methods that are largely based on idealized assumptions (e.g. normal distribution) cannot be used for statistical inference. This was shown by means of a simple example: The empirical p value was calculated for the difference between the scores of a treatment and a control group by using a resampling method called ‘Permutation’. A pdf file of the presentation can be downloaded here: Second_TRUG_meeting

First TRUG meeting

A sunny sky and a not-too-fresh breeze marked the beginning of the Twente R User Group. While enjoying drinks and tapas in the lush Twente countryside, both students and staff from several departments of the University of Twente met and exchanged experiences and ambitions regarding their use of the statistical programming software R.

The level of experience and type of ambitions varied greatly. Some hardcore programmers want to familiarize themselves with R to broaden their horizon, some use R for Monte Carlo simulation, some merely want to use a platform for data manipulation and analysis that does not require an expensive license, and some use R because it can do types of analyses that can’t be carried out using SPSS. Several members in the group were interested in parallelization and integration with C++ code to speed up analyses.

Martin Schmettow was the first to present. Martin, working at the department of Cognitive Psychology and Ergonomics, presented how to deal with card sorting data, showing some code how to construct graph data from the card sorting data (represented as lists of lists) using a recursive function, and how to best present analysis results: hot-spot plots rather than dendograms. Several libraries and their functions were discussed, like for instance plyr. Powerpoint slides can be found here: Cardsorting Workflow.

The User Group will meet monthly, alternating days and times so that more people are able to join, even if they can only attend bi-monthly. Next presenter, July 11, will be Inga Schwabe, from the Department of Research Methodology, Measurement and Data Analysis. July 11, 1700 hrs, Weleweg 410, Boekelo.