Statistical analysis is usually necessary to answer questions with behavioural data. Analysis should be planned and registered before collecting data. Once collected, a dataset should be formatted and permanently archived prior to analysis. Data is checked and visualised with descriptive statistics and graphs. Models representing hypotheses about the true effects present in the population from which the dataset is a sample are built and tested with inferential statistics. Many different hypotheses can be captured using a linear modelling framework in which an outcome variable is predicted with a combination of predictor variables and interactions. Sources of non-independence in datasets can be addressed with mixed models. The robustness of findings can be examined by comparing the results obtained when analysis is done in different ways using model selection and multiverse approaches. Confirmatory analysis designed to test preregistered hypotheses should be clearly differentiated from exploratory analysis that generates new hypotheses.

This chapter deals with the topic of feature expansion and imputation with a particular emphasis on computable features. While poor domain modelling may result in too many features being added to the model, there are times when plenty of value can be gained by looking into generating features from existing ones. The excess features can then be removed using feature selection techniques (discussed in the next chapter). Computable Features will be particularly useful if we know the underlining ML model is unable to do certain operations over the features, like multiplying them (e.g., if the ML involves a simple, linear modelling). Another type of feature expansion involves calculating a best effort approximation of values missing in the data (Feature Imputation). The most straightforward expansion for features happens when the raw data contains multiple items of information under a single column (Decomposing Complex Features). The chapter concludes by borrowing ideas from a technique used in SVMs called the kernel trick. The type of projections that practitioners have found useful can lend themselves to be applied directly without the use of kernels.