Comparisons
By way of comparison, 16PF-5 has only one scale whose alpha exceeds 0.80, and three
of its sixteen scales have alphas less than 0.70. OPQ Concept 5.2 shows eighteen
scales whose alphas are less than 0.70 and one is as low as 0.39! NEO-PIR has thirteen
scales whose alphas are lower than 0.70, and 16PF form A shows very few scales whose
alphas exceed 0.70. Remember that every scale of UKPQ has an alpha that exceeds
0.70 and four that exceed 0.80…
Over 15 months, OPQ Concept 5.2 showed much lower stability, coefficients ranging
from 0.73 to 0.48 with a median of 0.64. Similarly 16PF-5 over a much shorter period
showed coefficients of only 0.79 to 0.56
Remember UKPQ was 0.90 to 0.70!
UKPQ Reliability
The UKPQ's reliability was of paramount interest to the test designers.
Internal
consistency and test-retest reliability (or stability) are the two measures most
commonly used.
Click here to read about UKPQ’s superb validity.
Internal consistency
If all of the questions are part of the same personality construct,
then they should all be inter-correlated with each other. The average inter-correlation
of all the items in a scale, together with the number of items in that scale, is
used to calculate its coefficient alpha. The UKPQ shows four scales where alphas
exceed 0.8 and there are a further five where the alphas exceed 0.73. The remainder
of its seventeen scales have alphas that exceed 0.70 This is exceptional, and shows
that the UKPQ is exceedingly reliable.
Test-Retest Reliability (Stability)
If the UKPQ is measuring personality constructs
that will make a difference in the workplace, it is important that they be consistent
over a period of time. Indeed, we can argue that if a questionnaire is reliable,
then testing the same person at two different times should still produce broadly
the same results (if not exactly the same, because individuals do change over time).
The UKPQ shows exceptional test-retest stability. Results were correlated over an
average period of 12 months, and showed three scales with stability coefficients
that exceeded 0.9 and all other scales exceeding 0.70.