The article discusses the challenges of handling missing data in scientific research and presents two recommended approaches: maximum likelihood (ML) and Bayesian multiple imputation (MI). It clarifies misunderstandings about the missing at random (MAR) concept and criticizes older methods like case deletion and single imputation. The authors emphasize that missing data are not the focus of inquiry but must be handled carefully to avoid biased and inefficient results. They argue that ML and MI are currently the state of the art, with newer methods addressing non-MAR data. The article reviews the historical development of missing data methods, discusses the types and patterns of nonresponse, and explains the distribution of missingness. It also explores the plausibility of MAR, the implications of MAR, MCAR, and MNAR, and the performance of older methods like case deletion. The article concludes that while case deletion can be simple, it often leads to biased results, and that weighting and averaging available items are less effective. The authors recommend ML and MI as the best approaches for handling missing data.The article discusses the challenges of handling missing data in scientific research and presents two recommended approaches: maximum likelihood (ML) and Bayesian multiple imputation (MI). It clarifies misunderstandings about the missing at random (MAR) concept and criticizes older methods like case deletion and single imputation. The authors emphasize that missing data are not the focus of inquiry but must be handled carefully to avoid biased and inefficient results. They argue that ML and MI are currently the state of the art, with newer methods addressing non-MAR data. The article reviews the historical development of missing data methods, discusses the types and patterns of nonresponse, and explains the distribution of missingness. It also explores the plausibility of MAR, the implications of MAR, MCAR, and MNAR, and the performance of older methods like case deletion. The article concludes that while case deletion can be simple, it often leads to biased results, and that weighting and averaging available items are less effective. The authors recommend ML and MI as the best approaches for handling missing data.