The R package "mediation" provides comprehensive tools for causal mediation analysis in applied research. It supports two main approaches: model-based and design-based inference. The model-based approach estimates causal mediation effects and conducts sensitivity analysis under standard research designs, while the design-based approach offers tools applicable to different experimental designs with weaker assumptions. The package also handles multiple causally dependent mediators. The package allows users to investigate causal mechanisms using various data types and models, explore how results change with relaxed identification assumptions, and calculate quantities of interest under different research designs. It includes functions for estimating average causal mediation effects (ACME) and average direct effects (ADE), as well as conducting sensitivity analyses and moderated mediation. The package supports non-binary treatment variables, multilevel data analysis, and sensitivity analysis for sequential ignorability. The mediation package is available on CRAN and includes functions for mediation analysis, sensitivity analysis, and plotting. It supports a wide range of statistical models and allows for the analysis of multilevel data using lmer and glmer functions. The package also includes utilities for handling missing data through multiple imputation and provides tools for sensitivity analysis, including the medsens function. The package is designed to handle complex experimental designs, including single experiment, parallel, and parallel encouragement designs. It allows for the estimation of causal mediation effects under different assumptions and provides confidence intervals and p-values for the estimated effects. The package is freely available and has been updated to include new functionalities and improvements over the past three years.The R package "mediation" provides comprehensive tools for causal mediation analysis in applied research. It supports two main approaches: model-based and design-based inference. The model-based approach estimates causal mediation effects and conducts sensitivity analysis under standard research designs, while the design-based approach offers tools applicable to different experimental designs with weaker assumptions. The package also handles multiple causally dependent mediators. The package allows users to investigate causal mechanisms using various data types and models, explore how results change with relaxed identification assumptions, and calculate quantities of interest under different research designs. It includes functions for estimating average causal mediation effects (ACME) and average direct effects (ADE), as well as conducting sensitivity analyses and moderated mediation. The package supports non-binary treatment variables, multilevel data analysis, and sensitivity analysis for sequential ignorability. The mediation package is available on CRAN and includes functions for mediation analysis, sensitivity analysis, and plotting. It supports a wide range of statistical models and allows for the analysis of multilevel data using lmer and glmer functions. The package also includes utilities for handling missing data through multiple imputation and provides tools for sensitivity analysis, including the medsens function. The package is designed to handle complex experimental designs, including single experiment, parallel, and parallel encouragement designs. It allows for the estimation of causal mediation effects under different assumptions and provides confidence intervals and p-values for the estimated effects. The package is freely available and has been updated to include new functionalities and improvements over the past three years.