EEID 2024 Quantitative Training

Introduction to Workshop Tools

(Leah R. Johnson, Virginia Tech)

In this workshop, we are focusing on Estimating environmental response functions using Bayesian inference.

As mentioned earlier, environmental factors often effect disease transmission via their effects on the traits of pathogens, hosts, and vectors.

Thus we will be focusing on tools/resources that help us to

  • access trait data
  • build models of trait responses to environmental factors (with Temperature as our primary example)
  • perform inference for these using a Bayesian approach

Learning objectives

  1. We will give you a brief overview to the tools that we will use, and how to access trait data either through the web or using one of the tools.

  2. You will then complete a practical to give you a chance to install tools and practice an introductory task.

Tools for Bayesian Analysis

Although it is entirely possible to program a Bayesian analysis from scratch in your favorite programming language, most folks use tools designed for this purpose.

For example, one of the earliest was BUGS/WinBUGS, and other software, such as JAGS/rjags grew out of it.

In this workshop we’ll use two related tools that can be used through R:

  • nimble: Numerical Inference for statistical Models using Bayesian and Likelihood Estimation; uses almost the same syntax as BUGS, etc; flexible; runs in R.
  • bayesTPC: wraps nimble to enable relatively easy, single-call fitting of a variety of the standard trait-environment relationships with reasonable default priors, while allowing implementation of bespoke functions as well.

Both allow numerical approximation of posterior distributions (more later!).

Installing nimble and bayesTPC

Installing both packages will proceed in three steps (more details in the practical)

You will need to install these in this order to ensure that everything will work properly.

Accessing Trait Data

If you are doing experimental work, then maybe you are generating your own trait data.


However, often we want or need to find published trait data that we can use to parameterize our models.

Ideally, data produced by scientific studies should be FAIR:

  • Findable
  • Accessible
  • Interoperable
  • Reusable

Although data is becoming more FAIR, not every data generator follows these priciples, and it can be hard to find good data.

Sometime we find data directly from papers. For traits, we have a few additional places you can look:

  • Open Traits Network - mostly a listing of data sets, with a lot of different options, from Arthropods to Zooplankton. Not specific to disease systems.
  • BioTraits - primarily thermal traits, including things like movement traits across a variety of species in a unified format. Not specific to disease systems.
  • VecTraits - a variety of traits of arthropod vectors of diseases in humans, animals, and plants. Digitized records are currently primarily thermal traits.

VectorByte’s VecTraits Database

We will primarily use data from VecTraits for three reasons:

  • the database is specific to organisms involved in disease transmission
  • the database is designed on FAIR principles (with Interoperability being the key one here)
  • access to VecTraits is incorporated into bayesTPC making it straightforward to include it into an analysis workflow

We can explore data through the VecTraits Online Interface.

Practical

You will now move on to the hands-on portion of this section. You will:

  • get tools installed
  • choose a dataset you are interested in playing with through the VecTraits web interface
  • practice downloading your chosen dataset using bayesTPC
  • learn a bit about the downloaded format so you can choose the appropriate columns and visualize your chosen dataset.