predlimits(x, confLevel), x REAL scalar, vector or matrix with no MISSING elements, 0 < confLevel < 1 scalar |

You can use macro predlimits() to compute prediction limits for y = E(y|x) + epsilon or y = E(y | x1, x2 ...) + epsilon after running regress("y=x") or regress("y=x1+x2+..+xk"). These are limits on the a future value of y for specified values of the predictor variable or variables. predlimits(x, confLevel), where x is a REAL vector with length(x) = number of predictors (1 for simple linear regression), returns vector(lower,upper), where lower and upper are prediction limit with confidence level confLevel. Argument confLevel must be a REAL scalar between 0.5 and 1. When confLevel < .5, a warning message is printed and 1 - confLevel is used. Example: After regress("y=x1 + x2 + x3"), predlimits(vector(2,3,4), .95) returns vector(lower,upper) where lower and upper are the limits when x1=2, x2=3 and x3=4 You can use predlimits() to get limits for several values at once, returning hconcat(lower,upper), where lower and upper are vectors of limits for the various values. For simple linear regression, x should be a vector containing the values for which you want prediction limits. When there are k predictors, x should be a matrix with k columns, and lower and upper will have length(nrows(x)). Example: After regress("y = x1 + x2"), predlimits(vconcat(vector(1,2)', vector(2,1.5)',vector(3,3.2)'), .95) returns hconcat(lower,upper) where, for example, lower[2] and upper [2] are the limits when x1 = 2 and x2 = 1.5. See also estimlimits(), regpred(), glmpred().

Gary Oehlert 2003-01-15