A frequentist performs statistical inference by finding the best fit parameters. The Bayesian finds the distribution of the parameters assuming a prior. Frequentist inference can be regarded as a special case of Bayesian inference when the prior is a Dirac delta-function. Bayesian inference is well suited to online learning, an experimental design under which the model is continuously updated as new data arrives. Prediction, under Bayesian inference, is the conditional expectation of the predicted variable under the posterior distribution of the parameter.
Which model is better? How would your interpretation of which variables are the most important change between models? Would you arrive at different conclusions about the market signals if you picked, say, Model 1 versus Model 2? How would you eliminate some of the ambiguity resulting from this outcome of statistical inference?Bayesian inference is ideally suited to model selection because the model evidence effectively penalizes over-parameterized models.