[Date Index] [Thread Index] [Author Index]

Re: Problems with my first package: Statistics`Common`RegressionCommon`BestFitParameters instead of BestFitParameters

• To: mathgroup at smc.vnet.net
• Subject: [mg58177] Re: [mg58049] Problems with my first package: Statistics`Common`RegressionCommon`BestFitParameters instead of BestFitParameters
• From: Mathematica Ezine <ezine at omegaconsultinggroup.com>
• Date: Tue, 21 Jun 2005 06:02:51 -0400 (EDT)
• References: <200506170918.FAA07763@smc.vnet.net>
• Sender: owner-wri-mathgroup at wolfram.com

Without seeing the package, it's difficult to say for sure what the 
problem is. So I'll make an educated guess. It's a common problem when 
using the Statistics` packages. What I have here is also described at:

http://omegaconsultinggroup.com/Services/HTMLLinks/ezv2i03_4.html

You don't have the correct contexts in your BeginPackage command.

In[1]:=<<Statistics`NonlinearFit

This makes both NonlinearRegress and BestFitParameters available to the 
user.

In[2]:= Complement[{"NonlinearRegress", "BestFitParameters"}, Names[]]
Out[2]={}

So you may be tempted to use this in your package.

BeginPackage["MyPackage`", "Statistics`NonlinearFit`"]

This does load Statistics`NonlinearFit`.
It does make NonlinearRegress available to you package.
It does *not* make BestFitParameters available to you package.
Why?

In[3]:=Context[NonlinearRegress]
Out[3]=Statistics`NonlinearFit`

In[4]:=Context[BestFitParameters]
Out[4]=Statistics`Common`RegressionCommon`

Statistics`NonlinearFit` loads another package to get the definition of 
BestFitParameters, but it does not automatically make the context of 
that package available to your package.

The best analogy I've heard for this is: A friend of my friend is not 
necessarily my friend. So the correct thing to do is explicitly list 
the context of all symbols you want to use in your package:

BeginPackage["MyPackage`", "Statistics`NonlinearFit`", 
"Statistics`Common`RegressionCommon`"]

----------------------------------------------
Omega Consulting
The final answer to your Mathematica needs.
http://omegaconsultinggroup.com

On Jun 17, 2005, at 4:18 AM, frank wrote:

> Hello,
>
> thanks to some hints I got on this list, I have managed to write my
> first package.
>
> There was one problem with options, but I solved this with the help of
> the list archive: It wouldn't accept {verbosity->0} because it is
> something different than Schulerstuff`ConstrainedFit`Private`verbosity.
> But with the help of
>
> http://forums.wolfram.com/mathgroup/archive/1994/Apr/msg00041.html
>
> and a
>
> verbosity::usage="foo";
>
> this works now.  However, there remains one problem.  Within a function
> in the package that is exported, I do
>
> ExportedFunction[whatever_,options___]:=Block[{...},
> ...
> RegressResult=NonlinearRegress[...]
> ...
> Return[Switch[verbosity /. {options},
>       0, BestFitParameters /. RegressResult,
>       1, {NMinResultReport, RegressResult}
>       ]];
> ]
>
> With verbosity->0, everything is fine now, but with 1 I get:
> ...
> Statistics`Common`RegressionCommon`BestFitParameters -> {a -> 0.500455,
>     b -> 0.999833},
> ...
>
> where I expected simply
>
> BestFitParameters -> {a -> 0.500455, b -> 0.999833},
>
> Can you tell me, without me creating a minimal example, what I can do 
> to
> simplify the Return value?
>
> Regards, Frank
> -- 
> Frank Küster
> Inst. f. Biochemie der Univ. Zürich
> Debian Developer
>

• References:
  • Problems with my first package: Statistics`Common`RegressionCommon`BestFitParameters instead of BestFitParameters
    • From: frank <frank@kuesterei.ch>