Phenomena and processes taking place during welding are usually very complex and, for this reason, should be described using multivariate methods. The article discusses the methodological basis and selected application areas as regards the solving of welding problems using statistical multivariate methods. In addition, the article presents exemplary applications of the design of experiment, multiple regression analysis, cluster analysis, principal component analysis and logistic regression analysis. The application of multivariate analyses provides the possibility of performing the mathematical description of joining processes, which, after verification, could be used to forecast results of such processes, particularly in relation to the properties and fatigue service life of welded structures.