- Model of Process
- Model of Errors
- Process error (natural variability)
- Observational errors

- Simultaneous estimation of all parameters

Previous slide | Next slide | Back to the first slide | View Graphic Version |

Biologists generally do not have the luxury of refining an instrument to remove measurement errors. Fisheries biologists are further constrained by the imperative to analyze historical data. It is very difficult to “replicate” field experiments in ecological studies, and is impossible to revisit the past. Biological processes typically variable. Fisheries data are often inaccurate and the accuracy may vary over time. These problems make extraction of information from fisheries data very difficult.

We often take the “nearest number” from the scientific literature and substitute it for the value of a parameter (e.g. a growth rate) in some model. This process ignores the fact the the numerical value is itself a parameter estimate with some unknown error structure.

Integrated statistical modeling attempts to maximize the information obtainable from data of questionable quality by creating a consistent modeling and parameter estimation framework.

I’ll try to show how this works with 2 relatively simple applications to tag-recapture data.

For further insights into integrated statistical modeling, refer to http://www.island.net/~otter/