jobEQ statistics:: Standard Groups

Throughout its range of tests jobEQ uses the principle of standard groups in order to give a relative indication of where a person scores in comparison to others. This documents aims at explaining the principles behind the standard group and at making the statistical principles comprehensible.

What is a standard group?

In "jobEQ speak", a standard group indicates how a population will typically score for a parameter. For instance, if we take proactivity, the question is how the population's score will be distributed for this parameter. Once we know how a population is distributed, we can say that a person's score is low or high, relatively speaking, compared to that particular population. For instance, an absolute score of 69% on proactivity may be very high compared to typical scores in France, while it will just within the range of the standard group for the UK. This person will be seen as very proactive by the large majority of the French, while his proactivity will be considered almost "average" in anglo_saxon countries.

Australia:	||||||||||||||||||||||||||||||||||||||||||||||||||		The green line indicates the score of the individual, the red part of the bar indicates the standard group and the blue area is outside the standard group.
France:	||||||||||||||||||||||||||||||||||||||||||||||||||
UK:	||||||||||||||||||||||||||||||||||||||||||||||||||
US:	||||||||||||||||||||||||||||||||||||||||||||||||||

How are jobEQ's standard groups developed?

jobEQ's standard groups are calculated by taking the means of a sample of a group (e.g. a country as the UK), adding one standard deviation to this means to find the upper limit of the standard group, and subtracting the standard deviation to find the lower limit. If we presuppose that the population is approximately normal distributed, we know by definition that approximately 66% of the population will fall within that standard group, while 17% of the test population will score higher than the standard group and 17% will score lower.

so :   standard group = (means-stdev,means+stdev)

How important are these groups for Validity of test results?

Actually, they are not really important. They just help to give an idea of how people fill out the questionnaire. We use them to generate visual charts and/or textual explanations of a person's scores. When jobEQ questionnaires are used for taking decisions, these standard groups do not matter. The correct use for taking decision is comparing a person to the top performers holding a certain position (jobEQ's model of excellence technology).

Are these groups "statistically correct"?

Given the use we make of these groups, just to help to determine where persons score relative to their peers, it's not that important to be exactly right. We want to give a «good approximation» of the meaning of a person's score and being a few percents off is not important.

Still, even with relatively small standard groups we get a good approximation of a standard group for a culture. If you are into statistics, the example below will help to clarify this. (note: we picked these samples for educational purposes: these are NOT the specific standard groups used in our system.)

The statistics below compare the scores on proactivity (parameter OF1P in the iWAM questionnaire) between a french population (FR) from our public profiling database with a British population (UK) from the same database. Both populations are working populations (ages between 18 and 65) and are mainly white collar workers from different sectors (both public and private sector, from education to consulting or from secretary to top executive). Persons in both samples voluntarily filled out the questionnaire between October 2000 and May 2002. Most completed high school, close to 50% have a BA level degree. The population is almost evenly distributed between men and women. (Statistical analysis have shown that age and gender differences for iWAM tend to be smaller than cultural differences - see further research).

We *only* tested 238 persons for France and 329 persons for the UK. Still, the error margins on the mean scores shows that these groups are large enough: these cultures are so different that an eventual statistical error for the groups is much smaller than the cultural difference found. Statistically speaking, the difference between the 2 cultures thus is very significant (P<0,001). In other words, there is less than one chance in thousand that the French would be as proactive as the Brits.

If we would now repeat the same exercise for comparing US and Australia, again using data from jobEQ's public database as on 28th May 2002, we would see that the differences for proactivity are not significant. Yet again, the statistics show that the possible statistical error on these groups is quite small, even if due to sample size the statistical error margin on the Australian group is much larger (3.15%) than the error on the larger US group (1.06%). So we can safely assume that a bigger sample would allow us to come to the same conclusions, that is: that the Americans do not differ very much from the Australians in proactivity (nor from the Brits). The important cultural differences can be seen from a group of charts we prepared for that.

Notes:

Descriptive Statistics

	N	Mean	Std. Error Mean	Std. Deviation	Variance
OF1P	83	,55497	,02	,217033	,047

GO TO: < jobEQ sample page > / < jobEQ course index page > / < jobEQ Applications & Cases >

(c) 2000-2002, INTEGRAL PERSPECTIVES GROUP -- All Rights Reserved For more information, send an e-mail to jobEQ