In this paper, we considered discriminant analysis as a method by which we can classify objects into groups and explain the degree of relationship between groups and certain properties of objects, the so-called, predictors. We concentrated on the linear discriminant analysis (LDA), which aims to estimate the linear combination of predictors that best discriminates the affiliation of individual objects to a group. The advantage of LDA over other classification methods is that LDA allows a category-dependent variable, and the independent variable does not have to be dichotomous. In the first chapter, we analyzed the main conditions that must be met to obtain optimal analysis results: multivariate normality of predictors, homogeneity of the matrix of variance and covariance, and linear relationships of predictors within each group. We presented a derivation and testing of discriminant functions for group separation, and classification equations for classifying objects in groups. We also described three main types of linear discriminant analysis: direct, sequential, and stepwise analysis, and the criteria for assessing the significance. We explained the importance of discriminant functions with canonical variables and graphically presented the obtained separation of groups. In the second chapter, we applied the described analysis to a database of remote sensing data of five different crops. The database consists of 36 observations representing crops and 4 variables representing measurements obtained by remote sensing. The analysis showed that only one variable significantly affects the classification, but the overall obtained classification did not turn out to be significant. By transforming the predictor variables, we tried to improve the results of the classification, but without success. By performing a quadratic discriminant analysis, we found that the data set used, is not suitable for illustrating discriminant analysis. In the third chapter, we analyzed data on kernel properties of three different wheat varieties. The database consists of 210 observations representing wheat varieties, and seven variables representing the geometric properties of wheat kernel. By performing a linear discriminant analysis on this data set, we concluded that all predictor variables significantly contribute to the separation of groups and to classifying observations into the correct group. Finally, we plotted the separations of wheat cultivars and concluded that the separation of groups is mostly influenced by the kernel size.