# Tolerance interval for linear regression

I have two positively correlated variables, Q95 (independent, 95-percentile flow in cumecs from the previous summer) and LIFE (dependent, a score describing the health of some invertebrates). I would like to be able to predict, to some degree of confidence, what value of Q95 is required to attain a given LIFE. From what I have read online, what I need is a linear regression (which I'm quite happy with) with a tolerance interval (which is new to me). However, I can't find a good description (i.e., one I can understand) of how to do this. Ultimately, I'd like to be able to say, "A Q95 of x is required to achieve the target LIFE of y 90% of the time, with 95% certainty". Ideally I'd like to be able to perform these calculations in Excel. Any help would be greatly appreciated.

Edit 14/12/11: I've since discovered how to do this in R, but it doesn't look like it's possible to do with Excel, due to a lack of functions for noncentral t and chisq distributions.

reg <- lm(y~x)
lower <- regtol.int(reg, side = 1, alpha = 0.05, P = 0.90)\$'1-sided.lower'

-
My "normal interval workbook" at quantdec.com/envstats/software contains macros "Tol2," "NCTDist," and "NCTInv" to evaluate the non-central T and illustrates how to compute normal-theory tolerance intervals with it. See the [K Tables] sheet. –  whuber Dec 14 '11 at 21:45