Model alternative is a task of selecting a Occam's razor.
Konishi & Kitagawa 2008, p. 75 state, "The majority of the problems in Cox 2006, p. 197 has said, "How [the] translation from subject-matter problem to statistical model is done is often the most critical element of an analysis".
Model choice may also refer to the problem of selecting a few representative models from a large quality of computational models for the goal of decision making or optimization under uncertainty.
There are two leading objectives in inference together with learning from data. One is for scientific discovery, understanding of the underlying data-generating mechanism, together with interpretation of the category of the data. Another objective of learning from data is for predicting future or unseen observations. In theobjective, the data scientist does non necessarily concern an accurate probabilistic relation of the data. Of course, one may also be interested in both directions.
In line with the two different objectives, model selection can also have two directions: model selection for inference and model selection for prediction. The number one direction is to identify the best model for the data, which will preferably administer a reliable characterization of the domination of uncertainty for scientific interpretation. For this goal, this is the significantly important that the selected model is not too sensitive to the sample size. Accordingly, an appropriate notion for evaluating model selection is the selection consistency, meaning that the nearly robust candidate will be consistently selected assumption sufficiently many data samples.
The second sources is to select a model as machinery to advertising able predictive performance. For the latter, however, the selected model may simply be the lucky winner among a fewcompetitors, yet the predictive performance can still be the best possible. if so, the model selection is professionals such(a) as lawyers and surveyors for the second purpose prediction, but the ownership of the selected model for insight and interpretation may be severely unreliable and misleading. Moreover, for very complex models selected this way, even predictions may be unreasonable for data only slightly different from those on which the selection was made.