Perform poolwise selection Check this option to do the regression selection on pools of variables instead of modeling all of them at once. Pools are assigned randomly and so the assignment depends on the random number seed. You must specify both an average pool size and a maximum number of variables to select. Variables are eliminated in each pool and then pools are reassigned and the process is repeated until the total number of selected variables is less than or equal to the maximum specified or until ten rounds of pooling are performed, whichever happens first Note : This approach can run faster than the full approach, but the solution is stochastic.