Competing model fits, for which their associated algorithm
and covariance files have been saved,
can be ranked, AICc weighted, and their corresponding parameter
values and SEs averaged.
This procedure works only for nested models whose saved algorithm
files (*.alg) are associated with the same conditions for model
initialization, and therefore also have the same number of parameters.
Models can only differ in their parameter values, though the number
of free and fixed parameters can vary among competing models.
Be aware that the SE of a fixed parameter is considered to have
a value of zero (0).
Another requirement is that all competing models have an associated
covariance file (*.cov) that shares
its file name (except its suffix) with its algorithm
file name, e.g., ‘model_123.alg’ and ‘model_123.cov’.
Model averaging is conducted according to Burnham and Anderson
(2002), and averaged parameter values and their standard errors
are reported.
Smartstats© will guide you through the procedure for calculating
likelihood profiles.
The results of this procedure can be saved in a text file.
A key reference is:
Burnham, K.P. and D.R. Anderson. 2002. Model Selection and Multi-Model
Inference: A Practical Information Theoretic Approach. 2nd Edition.
Springer. [Burnham]
[Anderson]
[Springer]
|