Showing posts with label statistical methods. Show all posts
Showing posts with label statistical methods. Show all posts

Friday, 5 September 2014

Two Sodankyläläistä in Oulu (cont.)

After four days in the University of Oulu, the premises start to appear less and less hostile and labyrinth-like, and most of our brain capacity can thus be allocated to the modern statistical methods.

They were all gone

The topics covered during the summer school provide a valuable overview of the tools and techniques that any scientist who deals with data may need to implement someday.

The first day was mainly an introduction to the concepts of probability, likelihood and statistics, with a particular focus on the two possible paradigms (frequentist and bayesian) and their different philosophies. The second day, the lectures dealt with parametric modelling and statistical inferences, while the main topic of the third day was regression and curve fitting. Finally, measurement uncertainty analysis, Markov Chain Monte Carlo methods, time series analysis and spatial statistics were on the programme of the last two days.

The Secret Map

Every afternoon, a couple of hours were dedicated to practicals directly related to the examples considered during the lectures. These exercises were implemented in R. R is a free software programming language specially designed for statistics, which contains a great amount of useful packages. In addition, R enables to obtain nice and highly-tailorable plots.

TOO OPTIMISTIC PRIORS

About forty students were taking part to this summer school, with various backgrounds from biology to quantum physics. As not everybody knew one another, a "get-together" evening was organised by Exactus at some bar downtown on Wednesday evening, with a concert of the Groovesisters, a local duo (keyboard and drums). It turned out that, albeit statisticians, the organisers had set slightly too optimistic priors regarding the number of students who would join in. As only eleven of them showed up, and as about eighty free-drink tickets had been bought, let us say that we did not run short of beer...

It was slightly cloudy

After this pretty intense yet fruitful week, the homesickness has become unbearable, and soon will we be driving north to enjoy the colourful ruska and go picking lingonberries.

Wednesday, 3 September 2014

Two Sodankyläläistä in Oulu

This week, the Exactus doctoral programme of the University of Oulu Graduate School (also known as UniOGS) organises a summer school around modern statistical methods for natural scientists. Exactus coordinates the doctoral training of students preparing a PhD in so-called "exact sciences", namely astronomy, physics, chemistry, mathematics and statistics. Incidentally, two of the members of the SGO staff who are currently preparing their PhDs within the Exactus doctoral programme are taking part to this summer school.

Statistics can provide the scientists with highly-valuable tools and methods which may not seem absolutely straightforward at first glance to the neophyte. Yet, the most confusing aspects encountered during the first day turned out not to be mathematical ones but had rather to do with finding directions inside the university buildings.

Entering the R gate


A FOUR-STOREY BUILDING

Let us just recall a few key figures. Slightly more than 15,000 students are studying in the University of Oulu. Population of Tähtelä: about 60 people (during working hours, in the peak season). Characteristic dimensions and main facts of the Polaria main building in Sodankylä: 60 x 60 x 7 m, two storeys, two corridors, about thirty rooms. In comparison, the Linnanmaa campus main building is more of the order of 300 x 400 x 15 m, with four storeys, too many corridors and an infinity (or so) of rooms. No wonder why such a vast and overcrowded place traumatised those two Tähtelä guys far more than the Bayesian inductive learning approach (and its implementation in R) to predict the color of the next ball drawn from a box given the outcome of a first observation.

We were clearly outnumbered

Hopefully, by the end of the week, the probability to reach the next lecture room without getting lost given the location of the previous one will be of the same order as the probability of being eaten by mosquitos while wandering by the Kitinen river in July.

To be continued...