Page 142 - PhD GT

factor analysis, this distinction did not result in separate factors in the scale (see below). The adapted inventory was subjected to validation in a sample of 118 people attending the Leeds Addiction Unit for treatment of a heroin problem.
6.5.2 Procedures for assessing the reliability and validity of the adapted Coping Behaviours Inventory (CBIdrg)
The sample (Sample 4), the battery of questionnaires and the procedures for the validation of the CBI(drg) were described in Chapter 5, section 5.7.3 and 5.7.4. Table 5.1 in Chapter 5 presents a summary of the instruments administered. Complete questionnaire sets were obtained for 100 individuals and the adapted questionnaire was completed by 118 individuals. Test re-test reliability analysis was conducted with a separate sample of 61 respondents described as Sample 5 in Chapter 4. The CBI(drg) was administered to this sample on two occasions separated by a minimum of 2 days and a maximum of 14 days. Procedures and sample characteristics are described in Chapter 5, section 5.7.9.
6.5.3 Results: item analysis and internal consistency
The mean total score for the adapted Coping Behaviours Inventory was 53.7 with a range of 0 to 92. The respondent scoring 0 simply put a line through the whole questionnaire but it is not clear whether this was a reluctance to respond to individual items or the finding that none of these coping strategies were used. The distribution of total scores followed the normal distribution curve. Responses covered the full range of response choices and, with the exception of the respondent described above, all respondents used at least some of the items. Mean item score was 1.4 with the lowest mean item score being .23 (item 12) and the highest 2.26 (item 5). The mean inter-item correlation was .23. Item-total correlations ranged from .17 to .67 with the lowest item-total correlations being for behavioural coping strategies (items 33, 12, 8, 2, 37 and 10) and the highest item-total correlations being for cognitive coping strategies (items 32, 31, 19, 17, 22 and 13). The adapted scale had high internal consistency, with an alpha coefficient of .92. Alpha was not markedly raised by the deletion of any item (see Appendix 12).
Factor analysis was also used to assess reliability of the adapted scale as it takes account of the fact that items may measure a construct unequally (Carmines & Zeller 1979). Principal Components Analysis extracted 11 factors with an eigenvalue greater than one which accounted for 67% of the variance. The first factor had an eigenvalue of 10 and accounted for 26.4% of the variance. Thirty-two of the items had factor loadings over .35 on the first factor (Spector 1992 p. 55
136