# Lab 3

In Lab 3, we'll study the use of CDA models for estimation of some http://www.birdlife.org/datazone/species/terms/criteria.html" target="_blank">IUCN risk criteria, particularly criteria A2. We'll look at a number of different data sets, and look at what causes poors estimates and explore which risk metrics tend to be well measured (versus those that are poorly measured).

To run lab:

At the matlab prompt, you'll type 'Lab3'.

You'll be asked to choose a datatype, then you'll be asked to choose a quasi-extinction level, and then a time horizon.

Quasi-extinction level: 80% decline, 50% decline, or 20% decline. These correspond to the IUCN criteria A2 for critically endangered, endangered, and vunerable.

Time horizon: Choose the number of years for the extinction horizon. IUCN uses 10 years or 3 generations. For this lab, we'll look at 10, 25 and 50 year horizons (you can type in other horizons if you wish though).

What you will see:

Figure 1: Shows some diagnostics for the data and the posteriors for mu, s2p (process error), and s2np. The posterior is based on a uniform priors. How to interpret such a posterior for the purpose of scientific inference (e.g. assigning something equivalent to a P-value for frequentists) is up to considerable debate. Nonetheless, it represents a common currency by which we can compare the precision of CDA risk estimates for different data sets. In this lab, it is used in this sense, as a standardized metric of precision (without making any absolute inferential statements).

Figure 2: Shows the posteriors for some risk metrics. THESE RISK METRICS ASSUME THE s2np TERM IS MEASUREMENT ERROR, SO NOT RELEVENT FOR TRUE POPULATION RISK ESTIMATION (although it most definitely affects uncertainty).

How to interpret the posteriors for this lab.

Think about them as a relative measure of data support (relative to the other risk metrics and other. If most of the area of the posterior is within a particular range, that indicates that most of the data support is within that range.

You'll need generations times, and you'll find them in the left nav bar.

To run lab:

At the matlab prompt, you'll type 'Lab3'.

You'll be asked to choose a datatype, then you'll be asked to choose a quasi-extinction level, and then a time horizon.

Quasi-extinction level: 80% decline, 50% decline, or 20% decline. These correspond to the IUCN criteria A2 for critically endangered, endangered, and vunerable.

Time horizon: Choose the number of years for the extinction horizon. IUCN uses 10 years or 3 generations. For this lab, we'll look at 10, 25 and 50 year horizons (you can type in other horizons if you wish though).

What you will see:

Figure 1: Shows some diagnostics for the data and the posteriors for mu, s2p (process error), and s2np. The posterior is based on a uniform priors. How to interpret such a posterior for the purpose of scientific inference (e.g. assigning something equivalent to a P-value for frequentists) is up to considerable debate. Nonetheless, it represents a common currency by which we can compare the precision of CDA risk estimates for different data sets. In this lab, it is used in this sense, as a standardized metric of precision (without making any absolute inferential statements).

Figure 2: Shows the posteriors for some risk metrics. THESE RISK METRICS ASSUME THE s2np TERM IS MEASUREMENT ERROR, SO NOT RELEVENT FOR TRUE POPULATION RISK ESTIMATION (although it most definitely affects uncertainty).

How to interpret the posteriors for this lab.

Think about them as a relative measure of data support (relative to the other risk metrics and other. If most of the area of the posterior is within a particular range, that indicates that most of the data support is within that range.

You'll need generations times, and you'll find them in the left nav bar.

## No comments