Characteristics Value Estimate of Steel Bar Anchor Strength by Mean Response Confidence Interval

The statistical experiment design is an indispensable tool to the searching front the multivariable situations. Whichever the design statistical adopted or simply response of random variables, the technique of multiple regression for situations with normal distribution can be used to determine or to develop more adequate the statistical model. This model allows to infer on the experimental data. The estimates of the mean response of this model sufficiently are used in real applications of project, however in the structural applications the estimate of the response in the 5% down percentile is necessary, called characteristic value of the response. In this research, present the experimental strength results of bonded-in steel rods in timber in angle 45o between the rod and the direction of the grain, the estimates of the mean strength by analysis of multiple regression, the confidence interval of 90% and the characteristics response. The independent variables were used: anchorage area of steel bar surface, moisture of the wood and density. All combination of variables presented were six replicate. The wood used was Corymbia citriodora , dry to the air, with CCA treatment; were used galvanized threaded steel bars CA-50 and compound adhesive gel epoxy resin. This research results in the mean anchorage force, response of the analysis of multiple regression, the determination of the confidence interval and the estimate of the characteristic values for the variables combination, in the edges and inside of the interval of the experiment.