After generating a linear model, I often need to do subsequent operations to compute differences between contrasts and more accurately the difference between the coefficients of the fixed effects. Typically, I do this with the contrast package, but I also often need to compare multiple sets of contrasts which I can't do with contrast (as far as I know).
It would be helpful if I could provide a contrast matrix, a vector of coefficients (or fixed effects), and a variance-covariance matrix and use that to operate on the fixed effects of a model. I'd guess that this would be a new subclass of "errors" (something like a class of c("contrast_errors", "errors")) where the main "errors" object would work on the contrast values directly, and the "contrast_errors" object would work at the coefficient/fixed effects levels.
What do you think?
After generating a linear model, I often need to do subsequent operations to compute differences between contrasts and more accurately the difference between the coefficients of the fixed effects. Typically, I do this with the
contrastpackage, but I also often need to compare multiple sets of contrasts which I can't do withcontrast(as far as I know).It would be helpful if I could provide a contrast matrix, a vector of coefficients (or fixed effects), and a variance-covariance matrix and use that to operate on the fixed effects of a model. I'd guess that this would be a new subclass of "errors" (something like a class of c("contrast_errors", "errors")) where the main "errors" object would work on the contrast values directly, and the "contrast_errors" object would work at the coefficient/fixed effects levels.
What do you think?