CONFIDENCE INTERVALS FOR DISTANCE IN THE ANALYSIS OF VARIANCE

摘要: where y is the n x 1 column vector of observations, X k matrix xej's and 3 parameters. It assumed that model stated in such a way rank (this can always be done by eliminating any redundant parameters). The statistical problem to test null hypothesis, which specifies values first r parameters, ,81, . ., /, (r < k). Any hypothesis linear, independent, functions parameters put into this form suitable re-statement model. convenient partition X1 X2, consists columns X2 remaining k-r columns. P3 likewise partitioned 4, contains 4 Thus containing true we are interested, while treated as oT nuisance will also write yo for set parameter specified ' least squares estimate y, yT Y2 arbitrary points space, F, all possible sets values, If S1 sum testing yo, then