4.2 The role of parameters

In this section, models are characterized by how much information is needed to specify the model. The number of quantities (parameters) needed to do so gives some indication of how complex a model is, in the sense that many items are needed to describe a complex model. Arguments for a simple model include the following:

With few items required in its specification, it is more likely that each item can be determined more accurately.
It is more likely to be stable across time and across settings. That is, if the model does well today, it (perhaps with small changes to reflect inflation or similar phenomena) will probably do well tomorrow and will also do well in other, similar, situations.
Because data can often be irregular, a simple model may provide necessary smoothing.

Of course, complex models also have advantages.

With many items required in its specification, a complex model can more closely match reality.
With many items required in its specification, ...

Get Loss Models: From Data to Decisions, 4th Edition now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.