Page 209 - PhD GT

measuring those thoughts and behaviours which, in the main study were hypothesised to shed light on the nature of change in dependence. The Impaired Control Scale, described in Chapter Five, was used in the main study as a measure of concurrent validity for the measurement of dependence. In Principal Components Analysis, the five items forming the first part of this three part scale and named "attempts to control" (ICSAC), were found to form a separate component which was significantly negatively correlated with dependence. The remaining two parts of the scale were significantly positively correlated with dependence and thought to measure a component of the same phenomenon measured by the LDQ. The latter two parts of the Impaired Control Scale were not, therefore, included in the investigation of the independence of the five variables.
In order to test the independence of the variables to be used in the investigation of predictors of dependence, time 1 responses to all items in the Leeds Dependence Questionnaire, the General Health Questionnaire, the Social Satisfaction Questionnaire, the Coping Behaviours Inventory and the first five items of the Impaired Control Scale were subjected to Principal Components Analysis. Varimax Rotation with Kaiser Normalisation produced a five factor solution which clearly distinguished the five scales used (see Appendix 16). Establishing the independence of these five factors eliminated the problem of co-linearity and thus rendered them suitable for inclusion as potential predictor variables in regression analysis which is reported below.
9.4 Predictors of change in dependence
Having demonstrated that dependence as measured by the Leeds Dependence Questionnaire changed over time, the correlates of such change were examined; using bi-variate correlational analysis, the relationships between dependence scores, change in dependence at three months and other measures were described. These were reported in Section 9.2 above. In order to explore the predictors of change in dependence, regression analysis was used. This method of multi-variate analysis is designed to investigate the relationship between a single dependent variable and one or more independent variables. The model for regression analysis was constructed in the following way: the dependent variable, change in dependence was taken as a continuous measure in order to reflect the entire range of change that occurred. Scores for the change in dependence between intake and three months and between three months and twelve months were computed. Nine independent variables were chosen to be entered into the analysis. Three pre-treatment characteristics, namely age, sex and primary drug of misuse (heroin or alcohol) were included as the sample was not balanced in respect of these variables; five pretreatment functioning variables, shown above to be
204