Comments on: Fitting non-linear models

By: Dan

Dan — Sun, 07 Aug 2011 09:59:16 +0000

First of all I’m really fascinated by the possibility to do statistical work in Clojure/Lisp.

As I understand – maybe I am mistaken -, there is currently no function in Incanter for NLP optimiziation with (possibly) non-linear constraints. Is addition of such an algorithm planned ?

By: Chad

Chad — Wed, 26 May 2010 03:34:42 +0000

While it is nice to have the pretty picture that we see in the graph, this page is about fitting data. For the discussion to be complete I think one last command should be included. To obtain the parameters that fit the data we use the command:

(:coefs nlm1)

By: incanter

incanter — Wed, 10 Jun 2009 23:50:57 +0000

Another method for validation I use is to compare my results with certified results. NIST has data sets for which they have computed and certified estimates of model parameters and their associated statistics. So far I have validated Incanter’s results on linear and non-linear regression against these benchmarks.

The data sets and estimated parameters and statistics can be found here:
http://www.itl.nist.gov/div898/strd/general/dataarchive.html

By: incanter

incanter — Wed, 10 Jun 2009 19:40:23 +0000

Hi Matthew, I am indeed validating against R, which at this point makes sense since most of Incanter’s functions map, more or less, directly to R functions and not really to SAS or SPSS functions (yet). As I start developing higher level functions (e.g. EFA/CFA/SEM) I’ll start using other packages for validation (e.g. LISREL).

David

By: matthew

matthew — Wed, 10 Jun 2009 19:24:33 +0000

this is all pretty sweet… but are you benchmarking/validating your processes against established stats packages? i do assume that you’re checking against R but what about the other elephants in the room (i.e., SAS, SPSS)?