Modeling rape victimization in 4 Nairobi slums using discriminant analysis

Abstract

vi ABSTRACT Purpose: This study aimed at determining whether a girl’s Age, Slum of residency and Grade in school can be used as predictors of rape and also for classification of girls into different rape risk groups. Methods: Participants were a prospective cohort of 2383 adolescent girls 13-20 years old, attending one of 35 secondary schools selected by convenience sampling from 4 informal settlements of Nairobi, that is Huruma/Mathare, Dandora, Kibera, and Mukuru. These areas were selected because of their high rates of crime and nonparticipation in the previous studies conducted by Ujamaa Africa. Fisher's linear discriminant analysis (LDA) is a popular dataanalytic tool for studying the relationship between a set of predictors and a categorical response. LDA was used to analyze anonymously collected baseline data from the girls on 1) Incidences of rape 2) Slum the girls come from, 3) Ages of the girls and 4) The grade in school the girls are in. General linear model was also used to analyze the same variables to determine which of these contributes most to being sexually assaulted. Results: LDA prediction model created categorizes respondents into raped or not raped categories with 76.62% accuracy using Age, Slum and Grade as predictors of rape. The slum from where a girl comes from and the girl’s age contributes most to the likelihood of getting raped (age, slum p>0.001). 70% of girls are sexually assaulted by individuals that they know or are close to. Conclusion: LDA can use a combination of information on age, grade and the slum from where a girl comes from to categorize a girl into either the sexually assaulted or not groups with a 76.62% accuracy rate. The biggest contributors to the likelihood of being raped for girls are the slum from where a girl comes from, the girl’s age and the grade the girl is in at school. With respect to grade, girls in form 2 are the most targeted. Girls are mostly sexually assaulted by individuals close to them as opposed to strangers. Generally, sexual desirability, Empowerment/education and Economic status of a girl or woman can be used as predictors of sexual assault with respect to the likelihood of girls being raped or sexually assaulted