Re: [R] Regression coefficient constraints

2011-09-04 Thread Dimitris Rizopoulos
One option is to reparameterize to an unconstrained problem using the transformation: exp(betas_i) / sum(exp(betas_i)), with e.g., beta_1 = 0, and then use optim() to maximize with respect to betas. I hope it helps. Best, Dimitris On 9/4/2011 7:50 AM, andre bedon wrote: Hi Guys, Does any

[R] Regression coefficient constraints

2011-09-03 Thread andre bedon
Hi Guys, Does anyone know how I could constrain my regression coefficients so that they are positive and add up to one? Any help will be greatly appreciated. Kind Regards, Andre [[alternative HTML version deleted]]