Mickey et al. [Mickey, M.R., Dunn, O.J. and Clark, V., 1967, Note on use of stepwise regression in detecting outliers. Computers and Biomedical Research, 1, 105–111.] proposed a test for discordancy in linear models, which compares the sum of squares of residuals based on the complete model to that of a model obtained by deleting the potential outliers. John [John, J.A., 1978, Outliers in factorial experiments. Applied Statistics, 27, 111–119.] proposed a test procedure which treats the outliers as missing values and involves obtaining estimates of those missing values. In this article, we show that the two procedures are in fact equivalent. We also compare the performance of two methods of implementing these procedures. Owing to the wide variety of possible designs, tables of critical values are not readily available for testing for outliers in linear models. Therefore, we provide a program that will allow the user to input a design matrix, along with some data, and will output a p-value for testing for a specified number of outliers.