I'm not entirely sure I understand your question but perhaps the attached may help. Statistical significance is a poor criterion for selecting predictors. Look especially at our introduction.

Good luck with your work, David Booth

I used nonlinear least squares (nls) simulation to fit the coefficients of parameters, like the a, b, c in the following equation: y~a+b*x1+c*x2. The result indicates one coefficient is not significant. So I'm not sure if I could move forward to use the coefficients to do some prediction. Anybody could help?

Sen Yang. If you can't then you are in difficulty I'm afraid. See Gareth James, et al Introduction to Statistical Learning with R, and other references in our paper. BTW why can't you just replace our glm with your GLM and run the equivalent adaptive lasso. See our reference 22 for how to do that. All you need to do is download the R programs just like we did or I can send you our glm version and all you need to do in either case is change the . Note that your model is NOT nonlinear. Good luck. David Booth

David Eugene Booth I'm sorry I just gave a bad example. Actually the equation I try to simulate is a nonlinear equation. I want to fit the litter decomposition data with a exponential curve. Thanks so much for your reply.

Ok No problem with nonlinear models. Simply Google adaptive lasso for nonlinear models. I am less sure about software available but I suspect you can find some good R routines so why not try.? We know that p-value selection doesn't work so you're not likely to lose anything. May the Force Be With You. David Booth

David Eugene Booth May the force be with you! :)

Jochen Wilhelm

Jochen Wilhelm

Sen Yang The Very First thing you really need to do is write out your research question. Once your goals are known it becomes much easier to help you. Best wishes, David Booth