Standard Groups

What is a Standard Group?

A standard group is an indication of how a population will typically score on a test item. The indication is a range of typical scores. jobEQ uses this range on its feedback reports in order to give a relative indication of where a person scores in comparison to others. The standard group can be any group, such as a team of sales people, all employees of a certain organization, or quite often the population of a country.

Once we know how a group typically scores, 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 U.K. (once again, the percentages are relative scores). This person will be seen as very proactive by the large majority of the French, while his proactivity will be considered slightly above "average" in several other 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 helpful are these groups for adding validity to the test results?

The purpose of using standard groups is not to add statistical validity; rather standard groups help people understand the test results by showing how individuals compare to the rest of a population. We use them to generate visual charts and/or textual explanations of a person's scores. When jobEQ questionnaires are used for making decisions, these standard groups do not matter. The correct process for making decisions is comparing a person to the top performers holding a certain position (jobEQ's Model of Excellence technology).

How are jobEQ's Standard Groups developed?

Here is where it gets a little technical. jobEQ's standard groups are calculated by taking the means of a sample of a group (e.g. a country such as the United States), 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 two thirds of the population will fall within that standard group, while 1 out of 6 will score higher than the standard group and 1 out of 6 will score lower.

Many tests you see calibrate their standard group by testing student populations. This method, however, results in unrealistic results, so jobEQ used working-age participants (18 to 65 years old). The test participants used for the jobEQ standard groups have all been tested since 2000. Most completed high school, and most a white collar workers. The populations are evenly distributed between men and women. jobEQ continues to create standard groups for more countries as our client list expands across the globe. Of course, existing standard groups get updated as well.
Example: Documentation of the 2007 US Standardgroup

Are these groups statistically valid?

For those inquiring minds that want to know the statistics behind our standard groups: the error margin for our standard groups is always less than 5%. For Australia, it's 3.15%, for the U.K., it's 1.16%, and for the U.S. it's only 1.06% error. Once again, it is important to note that we use standard groups just as a guide to help understand test results, so the key is not determining the exact numbers; instead it is most important to get a close estimate that will illustrate how participants compare with their peers.

With these standard groups, we get a good approximation of the standard for a culture. 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 18 to 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).

We tested 238 persons for France and 329 persons for the U.K. We found that 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 two 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 British.

If we would now repeat the same exercise for comparing U.S. and Australia, again using data from jobEQ's public database as on May 28, 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 U.S. 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 British). The important cultural differences can be seen from a group of charts we prepared for that.

• Statistics generated using SPSS for Windows v.11 on an extraction of jobEQ's Public Profiling database as of May 28, 2002.
• The argument that the standard group for Australia wouldn't change much with sample size is confirmed by this second sample from July 11, 2002, after we started a call for Australian questionnaires. You'll notice that the change in mean is well within the predicted standard error of the first sample, and given this sample is a larger (n=83), the predicted error margin is smaller.

 
Back to Research, or continue to our Products