[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Certification value (was Bitwise operators)
I'm going to jump in here....
All of my comments are based on direct experience with the rigorous
certification test development process at Sun. Further, I'm also pretty
biased.
Scott, you directly mention "Has the correlation ever been
experimentally measured?" The answer is "Yes!", for every single version
of each Sun exam. (They are rewritten and retested from scratch every
few years.)
A bigger question is: What do the tests actually measure? They measure
different things (the whole point of having several different exams).
The SCJP exam tests knowledge of J2SE. Nothing more. It does not measure
how good a programmer you are, or what your work ethic is like, or if
you can meet a deadline or if you're a good person to have on a
development team. It simply tests knowledge of basic Java.
The SCJP exam is set up with the following in mind: If you've been a
Java programmer for a year, you should be able to walk in, sit down and
pass the exam. If you don't know Java, or you've only been using it
on-and-off for a little while, you should fail. That's how exam is
indended and structured.
This goal is acheived using a "beta" process: Hundreds of candidate
questions are written (on the order of 400 for the SCJP), and then many
people (who are anywhere from Java experts to completely unaware of
Java) sit down and take the whole exam (it takes about half a day). Each
individual question is then scored, based on the following criteria: The
Java experts got it right and the Java novices got it wrong. (There's a
whole, complex statistical model behind it, which I won't get into.) Any
question that does not score well based on these criteria is thrown out.
The rest are used to put together the actual exam.
Is this a perfect system? No; there are a number of flaws. First, the
exam is mutiple choice. Second, it is fairly easy to pick up one of the
study books and use that to pass the exam (because it is mutiple
choice). Third, people have taken the exam "shared" the answers with
friends (and the whole internet). These things actually devalue the
certification process because they allow a person to pass who should not
pass.
That's why there are many exams, both multiple-choice and practical.
While the SCJP exam will not answer the question "Can X write good Java
code?" (and it was never intended to), the combination of the SCJP and
SCJD exam will answer the question: "Can X use Java to solve a complex,
real-world problem?"
As for any sort of certification for "job performance", personally I
think that is at least 80% based on the actual work environment, not the
potential employee. Some people thrive in high-pressure start-ups,
others don't. No test can determine that.
Just my 2 cents.
- Paul Philion
Scott P. Smith wrote:
>I never said it was easy to pass all the various certifications tests. I
>did not mean to imply that. I probably wouldn't pass any one of them today.
>I would have to prepare for it a few weeks. If I passed, does it make me a
>better software engineer?
>
>I was commenting on the coloration between on the job performance and being
>certified.
>
>Has the correlation ever been experimentally measured? A test is not valid
>unless it is objectively correlated with the criteria it is attempting to
>measure. We take SATs to enter collage because the score on the SAT
>correlates highly with performance in collage. (Of course it's not
>perfect.)
>
>What you are tested for when you take the SCJP is performance on the SCJP.
>Extrapolating that to above average work performance is not necessarily
>valid. Maybe it is. I have a subjective opinion that I can't back up with
>proof. Unless it's backed up be evidence, It's a matter of faith. Either
>you believe in it or you don't. Maybe there is proof. But don't quote a
>study paid for by Sun. They are not objective. Is Gosling certified? I
>have no idea. If he is, did he actually sit down and take the tests.
>
>----- Original Message -----
>From: "Curt Smith" <chsmith@speakeasy.net>
>To: <ajug-members@www.ajug.org>
>Sent: Sunday, February 23, 2003 5:30 PM
>Subject: Re: Certification value (was Bitwise operators)
>
>
>
>
>>My incompetence at subtle sarcasm in my post below caused at least a few
>>folks to misunderstand that I was actually attempting (poorly) to advocate
>>certification. SCJP though would not be my suggested end point, rather
>>SCWCD plus SCEA. After passing both and many others, I found the
>>design of SCWCD to be an excellent test of ones knowledge of JSP, Servlet,
>>taglib, web.xml etc. Modern trick question design of multiple choice
>>
>>
>_can_
>
>
>>be tough and rigorous. Take the SCWCD if you think it's so easy and of no
>>value. :)
>>
>>Appologies, curt
>>
>>Curt Smith wrote:
>>
>>
>>
>>>>>I will probably offend many people by saying this, but I place no
>>>>>value in
>>>>>being certified.
>>>>>
>>>>>
>>>>
>>>>
>>>I agree regarding SCJP. Referring to beyond SCJP, analogously I'm
>>>glad the
>>>American Medical Assoc. doesn't feel that getting a BS in biology is
>>>sufficient
>>>to be an effective brain surgeon, there is also more to learn and is
>>>of essential
>>>value including and even beyond SCWCD, SCEA, IBM XML , , ,
>>>
>>>curt
>>>
>>>
>>>
>>--
>>
>>Curt Smith
>>chsmith@speakeasy.net
>>(h) 404-294-6686
>>(w) 404-463-0973
>>
>>
>>
>>
>>
>>
>>
>>
>
>
>