Re: Need for (FindFit, Refine) ?
- To: mathgroup at smc.vnet.net
- Subject: [mg42485] Re: Need for (FindFit, Refine) ?
- From: Bill Rowe <listuser at earthlink.net>
- Date: Thu, 10 Jul 2003 03:37:12 -0400 (EDT)
- Sender: owner-wri-mathgroup at wolfram.com
On 7/9/03 at 8:24 AM, ErsekTR at navair.navy.mil (Ersek, Ted R) wrote: > In Mathematica version 5 we have new functions FindFit, and Refine. Thanks for pointing these out. I hadn't noticed them before. > Why do we need FindFit when we already had Fit? The documentation indicates FindFit can do a few things Fit cannot do. It appears FindFit can handle non-linear problems as well as linear problems. Also, FindFit allows you to specify a different functions of the residuals to minimze. Fit only does a least squares fit. But you could minimize the absolute difference between the data and prediction with FindFit (the L1 norm). > What can we do with Refine that we couldn't previously do with Simplify or > FullSimplify? This isn't as obvious. The documentation indicates Refine is one of the things Simplify tries. So, if the goal is to have a simpler expression and you are used to Simplify you probably have little use for Refine. > If (FindFit and Refine) can do things we can't do with (Fit, Simplify, > FullSimplify) could they have just made the existing functions more > powerful? Possibly this could have been done. But the decision to add functions rather than modify existing fucntions most likely has to do with how things are coded in the source code. Perhaps, it was easier to add a new function than modify existing code and ensure current fuctionality would not be impacted. Certainly, by adding functions it is possible to guarantee no unintended side effects occur as is sometimes the case when modifying existing code. Note, it should go without saying I don't have access to the source code. So, the above is total speculation on my part.