Exact Simultaneous Confidence Bands in Multiple Linear Regression with Predictor Variables Constrained in Intervals

For the normal-error multiple linear regression model, the problem considered is the construction of a simultaneous confidence band with width proportional to the standard error for the regression function when the predictor variables are constrained in intervals. To date, exact band is available only when there is one predictor variable. When there are more than one predictor variables, only conservative bands are proposed in statistics literature. In this paper, a method is provided to construct an exact confidence band when there are any number of predictor variables. Also a criterion is proposed to assess the sensitivity of a simultaneous confidence band.