Page 122 - PhD GT

items all contribute to the measurement of that construct. The Impaired Control Scale has three parts which were shown, in the validation of the original questionnaire, to measure three separate though theoretically related aspects which contribute to an overall definition of impaired control. The first part is designed to address the question of whether the respondent has attempted control over their substance use while the next two parts examine the question of whether this control was achieved and whether the respondent believed that they would be able to assert control over use. Evidence to suggest that each of the sub-scales tapped a single construct was derived from principal components analysis of each part separately. Part 1 (attempts to control) yielded a single main factor accounting for 68.7% of the variance with all items loading above 0.77. Part 2 (failed control) yielded a two factor structure with the first factor accounting for 38% of the variance and a second factor accounting for a further 19% of the variance. All items loaded above .4 on the first factor and those items that were negatively worded had negative loadings greater than .4 on the second factor. Part 3 (perceived control) again yielded a two factor structure with a main factor accounting for 45.3% of the variance and a second factor accounting for a further 19% of the variance. Once again, the second factor was accounted for by the negatively worded items all of which had loadings greater than .5 on this factor. These findings are shown in Appendix 8; they are almost identical to those of Heather et al. (1998): in that study, a two factor solution was found for Parts 2 and 3 of the scale with the second factor accounting for 15% of the variance in each part and being made up of the negatively worded items which had large negative loadings on the factor.
Although the number of respondents who completed all three parts of the scale (n=111 as some of the data were missing in the case of 7 of the respondents) fell short of that recommended sample size for factor analysis (Nunnally 1978 recommends a subject item ratio of at least 10:1), principal components analysis of the whole scale was nonetheless carried out. The above findings were confirmed in that the analysis yielded three main factors, one was accounted for by loadings above .37 for all the items in Parts 2 and 3, the second factor was accounted for by loadings greater than .35 for negatively worded items in Parts 2 and 3 and the third factor was accounted for by item loadings greater than .4 for the first five items (the ICSAC sub-scale). These five items had high negative loadings on the first factor. In the rotated factor solution, the first five items had low negative loadings on the first and second factors and high positive loadings on the third factor. This analysis suggests that the five items making up attempts to control form one construct that is distinct from failed control and perceived failed control items, but does not distinguish the latter two parts of the scale from each other.
On the question of whether items based upon patterns of use, specifically items 3, 4, 9, 15, 117