"Boot Camp" in Wildlife Study Design and Data Analysis


BCSS has run 45 Boot Camps since 2009, located in 11 different countries and involving over 720 participants. You can see details on our Past Workshops page.

We target field researchers and decision-makers who know the biology, giving them a huge advantage over those with qualifications in statistics. They are not mathematicians, so we take a practical approach to problems with lots of activities and simulations.

What's different about wildlife?

Why hold a special workshop for wildlife study design and data analysis? What's different from normal biostatistics?

  • Our work is relevant to wildlife management, and we should produce results which are useful for decision-making ➜ Bayesian approach.
  • We deal directly with complex ecological systems with many interacting factors; we can't do experiments ➜ Modelling.
  • We collect binary data (present/absent, dead/alive,...) and count data, and have small samples; normality assumptions rarely apply ➜ binomial and Poisson distributions.
  • Our results are affected by the data collection process: we rarely detect all the animals or species present and must incorporate detection probability in the analysis ➜ specific designs and software (SPSS can't do it).

We focus on terrestrial vertebrates. Some of the techniques discussed can be adapted for invertebrates or plants, or for aquatic animals, but we have no expertise in those fields. Almost all our examples are for populations of terrestrial vertebrates.

What's the approach to learning?

Learners need to develop their own intuitive understanding of new concepts through exploration and experimentation, rather than absorbing explanations. So there are plenty of practical activities and a minimum of lecturing.

Throughout the workshop, we encourage participants to ask questions and discuss ideas with others, and we provide many opportunities for review and reinforcement.

Above all we believe that learning should be a positive experience - it should be fun!

What do we do during the Boot Camp?

We roll dice to simulate hearing frog calls and explore binary data, and we count orangutan in 1x1 km plots to understand the Poisson distribution. We explore these distributions in spreadsheet software.

Already on Day 1 we introduce R statistical software, and we use it throughout the workshop.

Everyone is screened for a disease, and we introduce Bayes Rule to make sense of the results - what's the probability of having the disease if you test positive? We then apply this to the orangutan counts to calculate an estimate and credible interval for the total number of orangutan.

Age is a continuous variable, and we use the ages of those in the room to talk about two-number summaries and alternative measures of centre and spread.

Applying probability to continuous variables means using probability density functions, and we use spinners to explore this idea. We also see how a normal distribution can result from adding up many small random effects.

We simulate trapping squirrels in the forest by drawing chips from a bag, and use Bayesian software to get a posterior distribution for the mean weight of squirrels in the forest. We see the typical output from Bayesian software and explore the effect of using different prior distributions. Then we apply these ideas to look at the difference in the size of crabs between areas with no fishing and areas with fishing.

Although our focus is Bayesian methods, we devote a day to maximum likelihood estimation (MLE) methods for estimating parameter values. This is widely used in wildlife data analysis, often as a preliminary way to explore the data before carrying out a Bayesian analysis.

We experiment with statistical models, and see how Akaike's Information Criterion (AIC) can help to select the best model for making predictions.

We generate binary data by throwing socks into boxes, noting our success and seeing if that is affected by the distance from the box or which hand we use. We develop a logistic regression model and see how maximum likelihood methods work in a spreadsheet. Then in R we try a range  of models and use AIC to compare them.

A comparison of MLE with Bayesian analysis for monitoring beluga whale populations shows how Bayesian results can be used for decision making.

We simulate surveys of marmosets to see (1) if the population is declining and (2) if marmosets need big trees. This leads to a discussion of survey design were we ask each participant to put up a research question for discussion.

The discussion includes sampling designs and the use of simulations in R to answer questions about sample size and optimal strategies.

We begin the exploration of occupancy concepts by going to a lawn and searching for ants. (We do have an indoor alternative in case of no lawns or wet weather.) We continue with the analysis in R of real data sets used in the seminal work on occupancy estimation. We use both MLE and Bayesian methods.

Spatially explicit capture-recapture (SECR) is introduced with simulations of geckos moving through habitat dotted with pit-fall traps. Although not all geckos are captured, the analysis usually gives a good estimate of the total number. Again this is followed up with MLE and Bayesian analysis of classic data sets and a discussion of survey design for SECR.

Long-term mark-recapture allows survival to be estimated. Although multi-year data sets are rare in the region, we show what can be done with them. We begin with an experiment simulating rat captures and also work with real data sets in R.

The final day is devoted to topics requested by participants, eg, review of key basic topics, more on advanced analyses, or discussion of participants' own projects.

What language will be used?

The Boot Camp is conducted in English. The aim is to provide participants with a starting point to follow up on their own the specific kinds of study design and data analysis needed for their own research. The resources for further study and the software manuals are all in English. So it's important to become familiar with the English terminology.

Sometimes we do pause and explain specific concepts in the local language if necessary.

Who should attend?

The workshop is aimed at science graduates who are involved in field-work in conservation or wildlife management, or who use the results of such field work. No previous knowledge of statistics is needed, ie, we'll assume you've forgotten the stats you learnt at university!

Participants should have a background in field biology, as that's where our examples come from.

We will assume familiarity with the use of computers - and in particular spreadsheets - and we'll ask you to bring a notebook computer to the course.

When and where?

The workshop covers ten days, with two one-day breaks, so 12 days in all.

See the BCSS home page for dates and venues of upcoming Boot Camps. If you want to help to organise a Boot Camp in your own country, please see here.


Page updated 18 May 2020 by Mike Meredith