Talk:Contrast (statistics)

Sum to Zero?
Not sure this is part of the definition of a contrast. It may be a prerequisite for validity in certain contexts, and that may be why the source states this. Japsai (talk) 15:47, 23 September 2008 (UTC)


 * I agree. Summing to zero is only required in the classical oneway anova context where the thetas represent groups means. In the more general contexts, where the thetas are arbitrary linear model coefficients, summing to zero may be neither required nor sensible. I have not seen the term contrast used in textbooks outside the classical context, but I have used term term myself in a general linear model context here: https://pubmed.ncbi.nlm.nih.gov/16646809/ GKSmyth (talk) 04:11, 28 May 2020 (UTC)