Monday, November 20, 2006

CentralTest International Tests

CentralTest International (CTI) has served more than a 6 million internet users throughout -
  • Spain

  • United Kingdom

  • France

  • Canada

  • United States

  • South-Africa

  • Australia

  • Gulf region

  • Europe

  • South America

  • Asia...

The CentralTest (CTI) tests are -


Read more about CentralTest International tests and clients .

3 comments:

Ian said...

I'm a registered psychologist with a PhD earned through validation and modelling of psychometric assessments and related theories. I would like to state that there is no such thing as a PhD-certified test. A PhD means something totally different. It's interesting that CentralTest have anything to do with such a high level qualification. If you read their test manuals and if you have a background in psychometrics/psychology, you'll see that they don't really understand what they are talking about. Reliabiity for one of their tests barely reaches the minimum acceptable level of .70.

Here's a couple more points from our extensive dossier on CentralTest:

Note very carefully, they do not report the Cronbach alphas at all – they merely phrase the sentence about them in an ambiguous manner – which could be taken as “all our alphas are above 0.70”, or “the APA threshold for acceptable alphas is 0.70”. Their statement at the bottom of page 5 is:

“The threshold value that assesses coherence between test scales is 0.70, and complies to
American Psychological Association recommendations”.

It is very strange they do not report the alphas, given they report the test-retest coefficients – which suggests there is something wrong with them!

Page 15: The Corporate IQ test

Ok – here we have a real problem – or at least they do!

The questionnaire consists of:

“Questionnaire : 38 questions including 12 logical questions, 11 numerical questions and
15 verbal questions”

And they note:

“Note : This test has been scientifically confirmed. Statistical studies were carried out on
group of over 500,000 web-surfers. The mean score was found to be very close to 100
and the standard deviation is 15. This distribution is representative of the entire
population.”

This is quite wrong. It reflects their lack of understanding of scaling of an IQ variable. What they have is a test consisting of 38 + 11 + 15 = 64 questions – scored right/wrong (as is the convention for items where you either get an item correct or incorrect). So the total test score range varies between 0 to 64, with a likely SD of around 10.

Now, you can either scale the entire sample of observed scores into an IQ variable score range with mean of 100 and SD of 15, which means the mean will be exactly 100 and the SD exactly 15, or
you express each observed score as it arrives, as a standard score, then re-express it as though it had come from a distribution with mean 100 and SD of 15. This will give you a mean which is likely to vary slightly around 100, and similarly for the SD, assuming the scores coming in also vary around their own initial mean and SD.

And there is the rub, what do you use as the initial mean and SD in order to express every future incoming score as a standard score, such that it can be re-expressed into the IQ variable metric?

If you re-scale the 500,000 web surfers, then their mean would be exactly 100 with SD of 15 – because that is how the scaling works.

So, again, even though they may not be cheating – one wonders whether they really understand what they are doing.

I further note no alphas are reported, nor any subscale correlations for the ability variables, let alone the other tests they describe. All in all, you and I know this level of reporting would fail every major country’s professional test body recommendations.

The ability tests will work after a fashion – but it’s anybody’s guess how well.

Having looked at their website and “team” – it’s no wonder they don’t know what they are doing! There isn’t a proper psychometrician amongst them!

Holland Codes Blog said...

Ian,

Thank you for providing that information. Do you have the information posted on a web site?

Mary

VijaiPandey said...

Hello Ian,

May I know a little more about you? You say you are a Ph.D holder and you earned your Ph.D on Validation and Modeling of psychometric assessment. Can you name your institute/ university?

Most of your comments on Central Test, my friend, are based on impression and not on data. Can you imagine of a company catering to 3000 odd companies (who must have employed thousands of qualified people in their HR department)worldwide can run successfully for long without scientifically valid and practically useful tools.

I do not know, but I guess you got very limited and filtered information on this. I do not want to comment on every line of your post as anyone can go to the website and verify himself/herself.

Just to bring to your knowledge, I am a qualified Psychometrician working in "TEAM" Central Test for last one year. And we have a "technical document" with detailed psychometric properties (including alpha for each scale)for all our tests.

and there exists the concept of Ph.D certified tools (may be with a different meaning than that of yours)just google a little and see yourself.