In the following section of the paper, I will first present reliability and validity checks that were required for particular measures employed in this study. I will then describe the way in which response latencies and error rates in the WrdTask were treated to obtain indices of implicit knowledge. In addressing the first set of hypotheses that involve relationships among measures, results from a correlational analysis, and a principal components analysis will be presented. To address the second set of hypotheses concerning whether the measures can be used to predict competence-related behaviour (rejection of influence), a regression analysis employing loadings obtained in the principal components analysis will be presented. Finally, to address the third hypothesis that exposure to two female subtypes will produce differential rates of rejection of influence, t-test results will be reported.
Before proceeding with the analyses, one case was excluded owing to the fact that the participant did not believe their partner was real. Responses on all measures were then examined for accuracy of data entry, and missing values. While values entered were in the range of responses allowed for by the scales, one missing value was found in the ASI data set. To replace this value, a group mean for this scale item was calculated.
The data were then checked for univariate outliers and assumptions of normality, and linearity. Eight univariate outliers were detected in the WrdTask response latency data and as such were removed. As kurtosis and skewness levels indicated a slightly peaked and positive distribution in the YrVws data, two outlying values were removed, thereby returning kurtosis and skewness to acceptable levels. In order to check for linearity, bivariate scatter plots were inspected. In many cases not a high degree of linearity was evident, however, there were no substantial nonlinear relationships apparent. Transformation of certain variables was considered, however, owing to the fact that this data set was combined with another data set (Foddy, 1998) later in the analysis, transformation may have made the results of this combined set difficult to interpret.