Mailing-List: contact dev-help@commons.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Commons Developers List" <dev@commons.apache.org>
Received-SPF: pass (athena.apache.org: domain of phil.steitz@gmail.com
 designates 209.85.218.43 as permitted sender)
Message-ID: <4E28DDE3.9000904@gmail.com>
Date: Thu, 21 Jul 2011 19:18:11 -0700
From: Phil Steitz <phil.steitz@gmail.com>
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6;
 rv:5.0) Gecko/20110624 Thunderbird/5.0
MIME-Version: 1.0
To: Commons Developers List <dev@commons.apache.org>
Subject: Re: [math] RegressionModelSpecificationException
References: <4E286678.4070709@gmail.com>
 <20110721181809.GD7045@dusk.harfang.homelinux.org>
 <4E28811F.4070508@gmail.com>
 <20110721202354.GE7045@dusk.harfang.homelinux.org>
 <CACMPRi2gYP75DG=jEzNDVVkX8q-VAEa4ZMgOLTeujeJcJ9Lw8w@mail.gmail.com>
In-Reply-To: 
 <CACMPRi2gYP75DG=jEzNDVVkX8q-VAEa4ZMgOLTeujeJcJ9Lw8w@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

On 7/21/11 6:53 PM, Greg Sterijevski wrote:
>>
>> +1 from me as well (does that even count?).

Thanks!  And yes, "that counts" - in fact it carries a lot of weight
because you have contributed more than anyone else to this area
recently.  We operate a "do-ocracy" here :)
> As an aside, I believe we will be adding more exceptions.

Yes.
>  The number of ways
> to screw up a regression is limited only by the user's imagination. As we
> finalize how we deal with obviously redundant regressors, that might be a
> new exception as well. Perhaps we should think of the exceptions cleaving in
> two planes. First would be input data problems. Second would be estimation
> failure.

That is a common pattern.  The first kind will extend
MathIllegalArgumentException (like
RegressionModelSpecificationException).  The second kind will depend
on what exactly is failing.  In some cases, it will be appropriate
to propagate an exception from one of the other packages; in others
we may want to define new exceptions specific to regression model
estimation.

Regarding obviously redundant regressors, it's probably in most
cases better to include this information in results, rather than
throwing exceptions, since in some cases users may want the
(arguably bogus) parameter estimates when we can compute them.

Phil
>
> -Greg
>


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org