The "Boot Camp" approach to learning stats

People should think, computers should work.

Computers - and computer software - are integral to statistical analysis today: all Boot Camp participants bring a laptop/notebook computer to the sessions.

Computers are good at arithmetic: they can be relied on to get their sums right, but we have to ensure that they are doing the right sums. Understanding the principles underlying the analysis we want to do is more important than ever if we are to avoid the temptation to accept whatever the "black box" churns out.

People should think

The first sessions focus on developing an intuitive understanding of sampling and making inferences from samples, including Information Theory and Bayesian analysis.

We use simple hands-on activities to simulate data: rolling dice to simulate detection of frog calls, pulling numbered chips from a bag to simulate sampling squirrels in a forest, shuffling clam shells from different sources to see if differences are real. We compare the results among participants, and compare too with the true value, which we know for our simulated data.

Comparisons work best with huge numbers of simulations: this can be done quickly on a computer, and R provides simple methods to do this.

These activities provide opportunities to explore concepts, discuss with other participants, and to ask all sorts of questions.

When we come to deal with the specific techniques used to estimate occupancy, population size, and diversity, we still use practical activities rolling dice, shuffling cards, searching for ants on the lawns - to simulate data for analysis. And of course we see examples of analysis of real-world data sets.

Computers should work

We make some use of LibreOffice Calc or Microsoft Excel for organising data and simple analyses, but most of the time we use the free statistical software package, R, and add-ons designed for wildlife data analysis.

R provides a uniform interface to functions for the analysis of presence/absence data or mark-recapture data for density or survival estimation. This enables us to focus on the principles of study design and the models underpinning the analysis.


Page updated 16 August 2014 by Mike Meredith