I-prior model selection can easily be done by comparing (marginal) likelihoods. We also explore a fully Bayesian alternative to model selection for linear models. A simulation study gives positive results for the I-prior for linear models in situations with multicollinearity, beating other types of priors such as g-priors, independent priors (ridge regression), and also Laplace/double exponential priors (Lasso).
Tests of Significance
For the regression model given in (1) and the function of the form
where each lies in a RKHS , one may be interested in which of these functions contribute significantly in explaining the response. A valid way of doing this is by performing inference on the scale parameters .
Where these parameters are estimated using maximum marginal likelihood, the distribution of the scale parameters are asymptotically normal. A straightforward test of significance can be done, and this would also correspond to the significance of the particular regression function since each is of the form
Model Comparison via Empirical Bayes Factors
Alternatively, model comparison can be done by comparing likelihoods. Again, as these models are fitted using maximum marginal likelihood, the log-likelihood ratio comparing two different models will follow an asymptotic distribution with degrees of freedom equal to the difference in the number of scale parameters in the models being compared. If the number of parameters are the same, then the model with the higher likelihood can be chosen.
Such a method of comparing marginal likelihoods can be seen as Bayesian model selection using empirical Bayes factors, where the Bayes factor of comparing model to model is defined as
The word ‘empirical’ stems from the fact that the parameters are estimated via an empirical Bayes approach (maximum marginal likelihood).
Bayesian Model Selection
When the number of predictors is large, then enumerating all possible model likelihoods for comparison becomes infeasible. Greedy selection methods such as forward or backward selection do exist, but one can never hope to completely explore the entire model space with these heuristic methods.
We employ a fully Bayesian treatment to explore the space of models and obtain posterior estimates for model probabilities given by
where is a model index and are the parameters of the model. The model with the highest probability can be chosen, or a quantity of interest estimated using model averaging techniques over a model set by way of
Variable Selection for Linear Models
For linear models of the form
Stochastic Search Methods
While posterior model probabilities may be enumerated one by one, variable selection with large would more than likely fail to list all probabilities. Various methods exist in the literature for stochastic search methods via Gibbs sampling, such that only models that have considerable probability of being selected are visited in the posterior MCMC chain.
One such method is by Kuo and Mallick (1998). Each model is indexed by , with a th value of zero indicating an exclusion, and one an inclusion of the variable . The above model (4) is estimated via Gibbs sampling, but with the mean of revised to incorporate :
This is done in conjunction with an I-prior on as in (5), and some suitable priors on , the intercept , scale parameters , and error precision .
Posterior inclusion probabilities for a particular variable can be estimated as , where is the th MCMC sample. This gives an indication of how often the variable was chosen in all possible models.
More importantly, posterior model probabilities may be estimated by calculating the proportion of a particular sequence , corresponding to a particular model , appearing in the MCMC samples.
A study was conducted to assess the performance of the I-prior, g-prior, independent prior and Lasso in choosing the correct variables across five different scenarios quantified by the signal to noise ratio (SNR). For each scenario, out of 100 variables with pairwise correlation of about 0.5, only were selected to form the “true” model and generate the responses according to the linear model above. The SNR as a percentage is defined as , and the five scenarios are made up of varying SNR from high to low: 90%, 75%, 50%, 25%, and 10%.
The experiment was conducted as follows:
- For each scenario, generate data .
- Obtain the highest probability model and count the number of false choices made by each method.
- Repeat 1-2 100 times to obtain averages.
- Repeat 1-3 for each of the five scenarios.
The results are tabulated below: