The limma package has a new function topTableF() which ranks genes by F-statistic. I am planning to add a feature to topTable() to access ranking by F-statistic. At present, topTable(fit, coef=c(2,3,4)) would be an error. In future, this will rank genes by the F-statistic for the 2nd, 3rd and 4th coefficients of fit. At present the default value for 'coef' is coef=1, so that topTable(fit) is the same as topTable(fit,coef=1) In future, 'coef' will default to coef=NULL, and the result will be a ranking by F-statistic for all the coefficients in the current model. If the fitted model only has one coefficient, then the result will be the same as at present. Will this change of behaviour cause anyone any serious problems? Does anyone feel that coef=1 should continue to be the default for topTable()? Thanks Gordon