About Add-my-Pet

Overview of the collection

The collection is complete for large phyla.
Chordates are complete at order level,
primates at family level.
A variety of related models captures animal life-cycle diversity and quantifies properties in parameters.

Although Dynamic Energy Budget (DEB) theory applies to all organisms, the AmP collection only deals with animals. The reason is that animals eat other organisms, which don't vary that much in chemical composition. Given the habitat, their environment can be characterized by the variables food availability and temperature as a first rough approximation. This characterisation is hard to make "complete" for other organisms, which hampers comparison. Comparison is the most useful asset of this collection.

To facilitate comparison, the estimated parameters are all expressed at the reference body temperature of 20 °C, even for endotherms and deep-sea or polar animals, which will probably be very dead at this temperature. So the temperature correction sometimes has academic elements, which needs to be remembered when it comes to making predictions. The extended Arrhenius model is used to account for effects of temperature: i.e. simple Arrhenius applies between a lower and an upper temperature boundary, but outside this tolerance range (dramatically) lower rates apply. In a few (e.g. polar) cases, where it was necessary to quantify rate reductions for increasing temperatures lower than the reference temperature, a different (i.e. lower) reference temperature had to be selected, which affects the comparability of species; the reference temperature needs to fall within the temperature tolerance range for consistency reasons. The extension is only used if enough data is available; the applied correction factor is given on the "implied traits" page, so it is always possible to undo the temperature correction.

At 2018/01/01, when the collection had 1000 entries, there were 112 zero-variate data types, and 153 univariate data types, in 588 combinations, written by 125 authors. So, few entries share the same data type combination and the number of data types is very much larger than the number of parameters. This argues for comparison on the basis of parameters, since all parameters were estimated for all species. Moreover, by being mechanistic (= based on first principles), DEB models interprete data, rather than just describe it, so can reveal inconsistencies in data and predict un-measured properties of species as functions of parameters.

Apart from extant species, the collection also has a number of extinct ones, demonstrating that DEB models can still be applied if data availability is poor. Examples are: giant skink, Deinosuchus, Pterodaustro, Tyrannosaurus, Archaeopteryx, great auk and Steller's sea cow. Needless to say, however: more data generally reduces uncertainty in parameter values.

Data completeness and mean relative errors

Survivor function of data completeness.
Survivor functions of the Mean Relative Error (MRE) and Symmetric Mean Squared Error (SMSE).
The relationship between mean relative
error and data completeness.
The relationship between symmetric mean squared error (SMSE) and mean relative error.