CAPÍTULO II: EN DEFENSA DE DERECHOS PERDIDOS
2.2. Breve crítica de la racionalidad liberal
2.2.3. Externalidades negativas y otros problemas tradicionales
Due to the difficulties of communication within the prison system it was not always possible to ascertain the reasons for non-attendance by all prisoners. It was also often unclear whether a prisoner had even received an invitation to attend the health care unit to take part in the research. A prisoner was therefore not included as someone who had refused unless a researcher had met with them, and the prisoner had declined participation at this point. In total, six prisoners declined to take part in the study. Of the six who refiised, two were known to have self-injured. Limited demographic details are available for those who refused to participate in Tables 1 and 2 of the Results section. Information relevant to the inclusion and exclusion criteria was not always available in the medical records and therefore seven prisoners were excluded
from the study at the point of meeting with a researcher. Three were excluded on the basis of psychotic symptoms, two were excluded due to language difficulties (English was not their first language) and two were excluded due to their self-injurious behaviour being solely suicidal in intent. All of those who were excluded on the basis of psychosis and language difficulties were known to have self-injured.
2.7 Design and Analysis of Data
Data was entered and analysed using the SPSS, statistical computer package. All data was then checked for accuracy prior to conducting any analyses.
This study was designed to investigate the association between a set of social rank- related variables in a sample of self-injurious male prisoners. This study attempted to replicate previous research that had found associations between shame, social comparisons, and submissive behaviour. In addition, this study also attempted to replicate findings that have identified an association between shame experience variables (shame, submissive behaviour, social comparisons) and anger; shame experience variables and childhood abuse; and childhood abuse and anger. In order to explore the relationships between the variables of interest, Pearson Product Moment Correlations were calculated for each association.
Furthermore, it was predicted that self-injurious prisoners would differ significantly from a group of controls on the above social rank-related variables. This study was therefore designed to investigate the effects of the between subjects independent variable, group (i.e. ISI group and control group) upon a set of dependent variables. These measures were grouped into three categories (i) measures assessing the
experience of shame (shame, social comparisons, submissive behaviour); (ii) measures assessing the level and expression of anger; and (iii) measures assessing childhood abuse and neglect. In order to explore these hypotheses t-tests and chi- square analyses were conducted to explore whether the two groups differed significantly on each dependent variable.
3. RESULTS
3.1 Data Analysis
Analysis of the data in this study required extensive use of parametric tests. One of the underlying assumptions of a parametric test is that any continuous dependent variable under investigation should be normally distributed. In order to assess the normality of the variable distributions, the significance of the skew for each variable was computed. This involved converting the skew values to z-scores, by dividing the skew value by the standard error of skew, z-scores less than 2.51 (p < 0.01) were considered normally distributed (Tabachnik & Fidell, 1996). For those variables which were significantly skewed, transformations were applied in an attempt to produce a normal distribution. The variables under investigation in this study were only ever positively skewed and therefore either square root or logarithmic transformations were applied. In order to allow for comparison with other studies the means presented in the tables below are based upon the original scores rather than the transformed scores.
The basis of a t-test also involves the comparison of two group means. The mean of a sample may however be unduly influenced by an extreme score, or outlier. Within this study, an outlier was defined as any score that deviated more than three standard deviations from the mean of that variable (Clark-Carter, 1997). Z-scores signify how far (in standard deviations) any one score is from the mean. All scores were therefore converted to z-scores and scores more than three standard deviations from the mean were thereby removed. The means displayed in the tables below are presented with the outliers removed in order to more accurately represent the data under analysis. It
is also important to note that this study involved distinct groups of participants and therefore normality and outliers were investigated separately for each group.
3.2 Participation
The total sample of recruited prisoners (Recruited Group) were compared with those prisoners who had attended for a meeting with the researcher but had declined to take part (Refusal Group). For those who refused to participate only limited demographic information was available including Age, Time served (in prison during this current detention). Ethnic origin and Prison status (i.e. Remand, Convicted but not sentenced. Sentenced). Analyses were therefore conducted to investigate whether the Recruited Group significantly differed form the Refusal Group on these specific variables. Table 1 and 2 display the comparative results for both groups.
Age and Time served were both continuous variables and as such independent samples t-tests were applied to examine whether the two groups differed on these variables. The Recruited group contained two outliers on the Time served variable and these were removed. The Time served variable also displayed a significant skew and therefore a logarithmic transformation was applied and a normally distributed variable was created. As can be seen from Table 1 below, the Refusal Group was significantly older than the Recruited group (t(7?) = 2.17, p = 0.033), but the groups did not differ on the Time Served in prison during this detention (t(75) = 0.57, p = ns).
The remaining variables of Ethnic Origin and Prison Status were both categorical. In general chi-square analyses would be applied to such categorical data, however due to the small number in the Refusal Group, more than 20% of cells in a chi-square
analyses would have contained fewer than 5 participants and thus any chi-square analysis would have been invalid. The Ethnic Origin variable was thereby collapsed into two categories, white and other. Likewise the Prison status variable was collapsed into sentenced and unsentenced. It was then possible to apply Fisher’s Exact Probability test to compare the two groups on these factors. As can be seen from Table 2, the groups did not differ significantly according to either Ethnic Origin (Fisher’s Exact p = ns) or Prison Status (Fisher’s Exact p = ns).
Table 1: Age and Time Served Demographics: A Comparison of the Refusal
and Recruited Samples
Demographics Recruited Group Refusal Group t df P
n mean (SD) n mean (SD) Age (years) 73 30.9 (7.08) 6 37.50 (8.50) 2.17 77 0.033* Time served (in months) 71 4.62 (6.32) 6 3.92 (3.11) 0.57 75 0.570 * p < 0.05
Table 2: Ethnic Origin and Prison Status Demographics: A Comparison of
the Refusal and Recruited Samples
Demographics Recruited Group Refusal Group Fisher’s Exact p
(n =73) (n = 6) n ( % ) n ( % ) Ethnic Origin 0.958 White 63 (6.85) 4 (16.67) Other 10 (2.74) 2 (16.67) .683 Prison Status Unsentenced 34 (32.88) 2 (16.67) Sentenced 39 (53.42) 4 (66.66) 77