.NET Opensource

Hello

I'm looking for a non-linear multiple regression statistics model...

From svn I see that it's not available by now... 

 do you have any good news about it for the future ?

 
Kind Regards!
 

Hi Lorenz,

Although there are plans for various regression algorithms, and someone even offered to contribute some code in this area some time ago (years actually; I haven't heard from him since): Unfortunately not, at least not for the near future. I just don't have the time at the moment...

On the other hand, I'm working on another project that's somewhat related to this topic where I might need this functionality myself soon, but I can't guarantee anything right now.

Actually, the problem is not linear versus non-linear; it's convex vs non-convex. There is quite an excellent book on the subject freely available at

http://www.stanford.edu/~boyd/cvxbook/

I might implement a generic convex optimization algorithm for Math.NET (yet, I have no schedule for that at this point).

Best regards,
Joannes Vermorel

http://www.lokad.com
 

I am curious to know if there has been any progress on a implementation of multiple linear regression within the library. I have created my own using Math.NET for one of my projects, however I am not convinced that it is the best solution (ie bugs/performace). So if the functionality exists in the library I would rather stick with it.

Keep up the good work,

Chris

I am also interested to find out if there are any plans for implementation of linear or non-linear regression in the Iridium libraries. I am fairly new to these concepts (mathematically) but I am willing to contribute in implementation, testing and documentation, where possible.

Regards,

Ivan

 Hi, I was also wondering if there is any regression development happening (similar to http://www.extremeoptimization.com/quickstart/PolynomialRegressionCS.aspx).

Thanks

There is support for polynomial regression currently in Iridium - see:

 http://community.opensourcedotnet.info/forums/p/574/56186.aspx#56186

 What I think is missing is non-linear regression (i.e. Levenberg–Marquardt algorithm) functionality.

I see. Is there a way to quickly get standard errors, confidence intervals, t-stats / p-values, etc.?

For example:

data / vairables: height, shoesize, isMale

If my (polynomial) empirical specification is

Expected Height = beta_0 + beta_1 * shoesize + beta_2 * isMale

I'm interested in getting the values of beta_0, beta_1, and beta_2 -- but also the standard error, t-statistics, and p-values of the estimates for those values. I usually use a stats program like STATA, but the scripting support is not great...

In that code example you gave me, I'm not sure how I'd quickly get those values. Apologies if there's an obvious answer to this, I'm not well versed!

Thanks.

Hello all, hello Christoph,

I would like to ask if there is anything new concerning the integration of multiple linear regression into your component.

I am looking for a solution that may be integrated into .net and I hope that I find it here :-)

Thanks for any feedback !

Regards,
Fabianus

hello everyone. i was also wondering if there has been any progress on this? it seems there is support for it but it also seems that getting the values aeblank mentioned is non-trivial. can anyone with experience in this area post a code snippet describing how to get the values from aeblank's post... the t stats and p vals? that would be most helpful.

~BG ~

Hi,

Does anyone know of a code snippet to find the values I mentioned in my last post?

Thanks.