The aim was to ascertain the test-retest reliability of a wide range of addictive and other adolescent risk behaviours using questionnaires in Swahili in Dar es Salaam, Xhosa in Cape Town, and Shona in Harare. The sample consisted of 119, 104 and 112 in the three sites respectively. Of the 43 variables for which Cohen's kappa was the appropriate measure, the reliability was poor for 0 (0.0%), fair for 5 (11.6%), moderate for 16 (37.2%), substantial for 20 (46.58%) and almost perfect for 2 (4.7%). Of the 27 variables for which the observed agreement was the most appropriate measure, it was above 90% for 22 (81.5%) variables. All three versions of our questionnaire provide data that is of acceptable reliability.
Human Sciences Research Council (Sat,) studied this question.