fitlm - only positive coefficients

Hi

I made a linear regression Model with fitlm. Some of the calculated coefficients are negative. In my Project this doesnt make sense. Is there an opportunity to set Boundaries for the coefficients or only allow positive ones?

Thank you!

Answer 1

[Torsten]

Use "lsqlin" instead. Here, you can set upper and lower bounds for the parameters.

[Tobias Grempels]

Here i need some help. How can i use lsqlin for optimizing my r-squared?

[Torsten]

The r-squared is only optimized in the model with no bounds on the regression coefficients.

You can compute the r-squared after fitting with lsqlin and with the bound constraints on the parameters, but it will of course be smaller than the optimal one.