Using stata IC v14.
my model theoretically mandates that there is no constant.
y= b1x1 +b2x2 ...+error
when the values of all predictors zero, y must be zero
i need a way of choosing a subset of predictors,GIVEN THIS CONSTRAINT.
i looked at the Vselect module available on ssc, but I don't thiink it has a Noconstant option.
would using adrian mander's LARS module to implement a lasso regression do what I need? As I understand it, Lasso done through this Lars will recalculate mean centered y and x, and the lasso output does not include a constant, but not sure if this achieveS my goal.
Any advice?
Thanks
my model theoretically mandates that there is no constant.
y= b1x1 +b2x2 ...+error
when the values of all predictors zero, y must be zero
i need a way of choosing a subset of predictors,GIVEN THIS CONSTRAINT.
i looked at the Vselect module available on ssc, but I don't thiink it has a Noconstant option.
would using adrian mander's LARS module to implement a lasso regression do what I need? As I understand it, Lasso done through this Lars will recalculate mean centered y and x, and the lasso output does not include a constant, but not sure if this achieveS my goal.
Any advice?
Thanks