Hi, I would like to understand why the residual standard error, and the degrees of freedom are changing when I define custom contrasts, which are not orthogonal. For example: y <- rnorm(40) x <- factor(rep(1:10,4)) summary(lm(y~x)) #standard model: Residual standard error: 1.103 on 30 degrees of freedom summary(lm(y~C(x,contr.sum))) #orthogonal contrasts: Residual standard error: 1.103 on 30 degrees of freedom summary(lm(y~C(x,c(rep(0.2,5),1,rep(0,4)),1))) #custom contrast: Residual standard error: 1.049 on 38 degrees of freedom In addition, is there a way to run the custom contrasts analysis without equal variance assumption? I expect the degrees of freedom to get lower in this case, as in ANOVA welch. Thanks! -- View this message in context: http://r.789695.n4.nabble.com/degrees-of-freedom-in-custom-contrasts-ananlysis-tp4673552.html Sent from the R help mailing list archive at Nabble.com.