This chapter provides a comprehensive review of global sensitivity analysis (SA) methods for model output. It distinguishes three main types of methods: screening, importance measures, and deep exploration of model behavior. Screening methods, such as the Morris method, are used to identify non-influential inputs among a large number of variables. Importance measures, including Pearson correlation coefficients, standard regression coefficients, and partial correlation coefficients, quantify the influence of each input on the output. Deep exploration methods, such as graphical and smoothing techniques, provide a detailed visualization of how the output varies with each input. The chapter also discusses metamodel-based methods, which use surrogate models to estimate sensitivity indices more efficiently. A synthesis of these methods is presented, highlighting their cost, complexity, and the type of information they provide. The chapter concludes with a discussion of open problems in SA, such as estimating total Sobol' indices at low cost and handling dependent inputs.This chapter provides a comprehensive review of global sensitivity analysis (SA) methods for model output. It distinguishes three main types of methods: screening, importance measures, and deep exploration of model behavior. Screening methods, such as the Morris method, are used to identify non-influential inputs among a large number of variables. Importance measures, including Pearson correlation coefficients, standard regression coefficients, and partial correlation coefficients, quantify the influence of each input on the output. Deep exploration methods, such as graphical and smoothing techniques, provide a detailed visualization of how the output varies with each input. The chapter also discusses metamodel-based methods, which use surrogate models to estimate sensitivity indices more efficiently. A synthesis of these methods is presented, highlighting their cost, complexity, and the type of information they provide. The chapter concludes with a discussion of open problems in SA, such as estimating total Sobol' indices at low cost and handling dependent inputs.