Intervention analysis: an example with data of the Lights4violence project in Alicante
Carmen Rodríguez Blázquez
M. João Forjaz
Belén Sanz Barbero
When planning to analyze the effect of an intervention, several questions arise: Does the outcome variable, measured in the same individual, show a change over time? Are the patterns of change different between individuals of the sample? What variables show variation within and between individuals? To answer these questions, a longitudinal design of repeated measures should be carried out. This type of study consists of analyzing the change over time in a sample of individuals, where the same outcome variable is measured in the same person before and after the intervention.
The intervention was the workshop with adolescents: "Filming Together to See Ourselves in a New Present". To perform this longitudinal analysis, data were collected from 85 students in Alicante, who answered the questionnaire before (first wave) and after (second wave) the workshop. In addition, data was drawn from 91 students who had not done the workshop, which is called as control group. For this group, we expect to find non-significant differences in the different scales.
The first analysis was to calculate the mean value of all scales in first (pre-test) and second (post-test) waves. To know if the difference between the means over time was significant, paired sample t-test was calculated. A preliminary comparison between means showed a significant decrease in the scales measuring violence, and significant increase in assertiveness and empathy in the intervention group. This change was significant in girls (Figure 1) and non-significant in boys (Figure 2). Non-significant differences were found in control group.
The mean differences between pre and post intervention were calculated without taking into account other variables that might be influencing the mean differences. Therefore, the mean values of the scales should be calculated adjusting for demographic and other variables related to violence experiences. In order to calculate the adjusted means, we used a generalized linear mixed model. For example, when using the adjusted means of assertiveness, we observed that the intervention group of girls increased their assertiveness while the boys remained stable:
*Means adjusted by time, age, mother and father studies, mother and father employ, violence dating, bullying/cyberbullying variables and violence experiences in childhood.
When we calculated the means adjusted by sociodemographic and violence variables, the assertiveness increased in the intervention group and slightly decreased in the control group:
*Means adjusted by time, sex, age, mother and father studies, mother and father employ, violence dating, bullying/cyberbullying variables and violence experiences in childhood.
ISCIII team is analyzing the intervention data and more results will follow in the next weeks.