How can I get higher rsqured of Multivariate Polynomial Regression?

Hi I used this function with my data and R-squared is 0.19 which is really low.How can I get higher R-squared. Should I use optimization ?

[Greg Heath]

What is "this function" ?

[Rita]

Multivariate Polynomial Regression

http://www.mathworks.com/matlabcentral/fileexchange/34918-multivariate-polynomial-regression

Answer 1

[John D'Errico]

Congratulations! You win the award for being the 10 million'th person to WANT a better fit than they got. Sadly, just wanting something is not sufficient. If it was, then the Cubs would have won a World Series long ago. If it was, then every person who ever bought a lottery ticket would have won the first prize. If it was, then we would have achieved peace in the mid-east long ago.

There is no magic to be found. Anyway, you can think of the regression estimate as the result of an optimization ALREADY. A further optimization won't help, because what you got is already the best you can get.

To get better results, you must get better data, with lower variability. You choose a better, more intelligent model, that better represents the data you do have.

[Rita]

Thanks John,I got higher R squared=0.76 with your function "polyfitn" but I got an error in estimation of polyvaln.

[John D'Errico]

That you got a higher R^2 probably means you chose a different model. It may not have been a good choice of model. You may be overfitting the data.

I don't know what error you got. polyvaln does no estimation, so I can't guess what you mean.

If you want, feel free to attach your data so I can take a look at it.