I have a distribution that looks almost linear (can be slightly concave or slightly convex) between two positive boundaries;
I tested many families in the gamlss package but cannot find anything that fits well here. The best solution I have is so far is to estimate the boundaries beforehand, and use trunc() to create truncated regressions (then using exponential distributions). However the boundaries vary with X, so I need to estimate them beforehand with a quantile regression, I'd like to know whether there exists another general family that could work better.