Note: You must be registered in order to post a reply. To register, click here. Registration is FREE!
T O P I C R E V I E W
Posted - 17 Jan 2008 : 00:11:34 I'm seeing two behaviors I don't understand:
1. Can there ever be a case when a SplitMulReg has a consider vector with a particular iv set to REG_VAROUT (-1), yet after the regression is solved, a coefficient is returned in that position? Shouldn't a coefficient of zero always be returned when consider is set to REG_VAROUT for that iv?
2. Is it ever correct behavior for a regression to modify the consider vector when a solution is run?
Thanks in advance -
1 L A T E S T R E P L I E S (Newest First)
Posted - 17 Jan 2008 : 11:38:50 1. If an element of the coefficient Consider vector for a split regression has been set to REG_VAROUT, the returned coefficient should be 0.0. If you are seeing different behavior, see if you can reproduce the problem using our example program (we can't). If you can't reproduce the problem with our example program, you need to supply us with a zip file of the complete source to the simplest possible program that reproduces the problem. You can send it firstname.lastname@example.org.
2. The software should modify the Consider vector if you set an element to REG_VARCONSIDER. It will modify it to be either REG_VARIN or REG_VAROUT, depending on whether the variable was left in the final solution, or left out. We see that in the case of split regression, the Consider vector is not being updated with the final solution flags (REG_VARIN or REG_VAROUT), thought the coefficients are correct, i.e. a non-zero coefficient means that a variable was included in the final solution. We will correct this in the next revision.