Shouldn't DurbinWatsonD be an options of NonlinearModelFit?
- To: mathgroup at smc.vnet.net
- Subject: [mg124890] Shouldn't DurbinWatsonD be an options of NonlinearModelFit?
- From: Gilmar Rodriguez-pierluissi <peacenova at yahoo.com>
- Date: Sat, 11 Feb 2012 06:34:27 -0500 (EST)
- Delivered-to: l-mathgroup@mail-archive0.wolfram.com
- Reply-to: Gilmar Rodriguez-pierluissi <peacenova at yahoo.com>
Back in May 2011 Peter Pein: http://forums.wolfram.com/mathgroup/archive/2011/May/msg00442.html raised an interesting question that I can't get out of my head: [In using DurbinWatsonD]"Are you sure you didn't want *Non*LinearModelFit (as x^2 is'nt linear at all)??" Shouldn't DurbinWatsonD be incorporated as part of the statistical options of NonlinearModelFit? If I evaluate: data={{217928.`,3546.664056`},{219129.`,3647.504482`},{221577.`,3712.326693`},{227481.`,3785.893753`},{231748.`,3832.600727`},{233514.`,3921.246847`},{232857.`,4072.872193`},{233664.`,4202.447982`},{235228.`,4193.074776`},{240526.`,4406.875387`},{243310.`,4520.20992`},{249587.`,4684.744763`},{250128.`,5046.968457`},{253383.`,5202.034503`},{257751.`,5388.147672`},{261999.`,5487.223742`},{258229.`,5573.48027`},{262567.`,5761.257796`},{263272.`,5801.166528`},{267643.`,5870.756827`},{272468.`,6015.724222`},{274035.`,6080.340626`},{276154.`,6156.352757`},{278323.`,6381.948988`},{282606.`,6734.369932`},{289505.`,6946.336039`},{295243.`,7368.445402`},{293956.`,7505.902557`},{294410.`,7773.3562680000005`},{296399.`,8043.178193`},{297382.`,8172.984975`},{298289.`,8294.955802`},{299248.`,8599.118621`},{296785.`,8844.871`},{299359.`,9324.551464`},{300184.`,9488.654185`},{299993.`,9447.234793`}}; nlm=NonlinearModelFit[data,a +b x+c x^2+d x^3,{a,b,c,d},x] Normal[nlm] -211581.+2.54863 x-0.0000101522 x2+1.3686×10-11 x3 LinearModelFit[poprpi19722008,{1,x^2,x^3},x]["DurbinWatsonD"] Gives: 0.246693 But; NonlinearModelFit[poprpi19722008,{1,x^2,x^3},{a,b,c,d},x]["DurbinWatsonD"] gives a spurious output. Thank you! Gilmar Rodriguez Pierluissi