Re: Question about DurbinWatsonD
- To: mathgroup at smc.vnet.net
- Subject: [mg119071] Re: Question about DurbinWatsonD
- From: Darren Glosemeyer <darreng at wolfram.com>
- Date: Sat, 21 May 2011 06:46:40 -0400 (EDT)
On 5/20/2011 5:37 AM, Gilmar Rodriguez-pierluissi wrote: > In: > > http://reference.wolfram.com/mathematica/RegressionCommon/ref/DurbinWatsonD.html > > it says that: > > "As of Version 7.0, DurbinWatsonD has become a property of LinearModelFit". > > But, when I evaluate: > > data = {{0.05, 90}, {0.09, 95}, {0.14, 110}, {0.17, 125}, {0.2, 140}, {0.21, 150}, {0.23, 175}, {0.25, 190}, {0.3, 210}, {0.35, 255}}; > > LinearModelFit[data, {1, x^2}, x, RegressionReport -> {DurbinWatsonD}] > > the above line will not produce the DurbinWatsonD value of 1.14 shown in the example on that web page. > > How can I then get the DurbinWatsonD value? > > Thank you! > > Gilmar Rodriguez Pierluissi > LinearModelFit operates differently than the Regress function did. LinearModelFit generates a FittedModel object from which results such as "DurbinWatsonD" can be obtained, so you could do the following: data={{0.05, 90}, {0.09, 95}, {0.14, 110}, {0.17, 125}, {0.2, 140}, {0.21, 150}, {0.23, 175}, {0.25, 190}, {0.3, 210}, {0.35, 255}}; lm=LinearModelFit[data, {1, x^2}, x] lm["DurbinWatsonD"] The main advantage of this object-based approach over the way Regress operated is that you can get additional results after the fitting without having to re-fit the model. http://reference.wolfram.com/mathematica/ref/LinearModelFit.html contains a number of examples and the list of properties available for linear models. Darren Glosemeyer Wolfram Research