Which selection method creates a sequence of models of increasing complexity starting from the baseline model?

The selection method that creates a sequence of models of increasing complexity, starting from the baseline model, is the forward selection method. In this approach, the process begins with a simple model that includes no predictors, and then it gradually adds one predictor at a time based on a specific criterion such as statistical significance or improvement in model fit. The goal is to evaluate the impact of each added variable on the model, ensuring that every step improves the model's performance, thereby building complexity progressively.

This method is particularly useful when there is a large set of potential predictors, as it helps in identifying which variables contribute meaningful information to the model. Additionally, by starting with a baseline model, forward selection allows for a clear understanding of how each added variable enhances the model's predictive capabilities. As complexity increases, so does the model's ability to fit the training data, which can lead to better insights and predictions when applied to new data.

In contrast, backward selection begins with a full model and removes predictors, while stepwise selection combines aspects of both forward and backward methods. Recursive selection is not a standard term used in the context of model selection methods in statistics. Thus, forward selection is the best choice for the scenario described.