Review of parametric_ci() #512
Description
Not so much bug as I don't think anything is wrong but I think parametric_ci() needs to be looked at as I think the current implementation makes a bunch of hidden assumptions that make it a likely source of future bugs. Main things that it assumes that for the p-value the null hypothesis is 0 and that the p/q functions have been standardised to return values on a SE=1 scale. Ideally the mean + SE information should be imbedded into the p/q functions and this function shouldn't do anything with regards to re-scaling. Though if adapting care is needed about the order of confidence intervals for negative values as I think the currently implementation would incorrectly flip them if the q/p functions had a negative mean.