Abstract:  I show that the likelihood developed for regression-to-the-mean models by Kazemnejad et. al. (2002) can be simplified so that closed form expressions for hypothesis tests and confidence intervals are obtained.  I show both that these formulas are asymptotically correct (when sample size goes to infinity) and that they work well for small samples.  I use the method to analyze an example data set.  

Key words: bivariate normal model, matched-pairs, model misspecification, nuisance parameter, regression to the mean