Aim:

To investigate whether pubertal timing has an impact on bullying and victimisation at school for males and females and whether the relationship between peer victimisation and mental health problems (such as depression symptoms, psychosis symptoms, anxiety or self-harming behaviour) would be exacerbated by early maturation (girls) or late maturation (boys).

Hypotheses:

1) Early-maturing girls will be more likely to be victimised.

Outcome Variables:

Bullying status groups according to child, mother and teacher report at ages 8 to 18.

Exposure Variables:

Puberty measures from 8 years to 17 years

Confounding variables:

Demographics: gender, ethnicity, family social and economic status, home ownership;

Family factors: domestic violence and harsh parenting, maternal mental health status, child abuse and maltreatment;

Psychological factors: any mental health problems using DAWBA, internalizing and externalizing problems using SDQ, depression using SMFQ and MFQ;

Individual factors: sex, IQ, sexual relationships

2) Late-maturing boys will be more likely to be victimised.

Outcome, exposure and confounding variables:

Listed as above.

3) Off-time puberty will mediate/moderate the relationship between bullying involvement (at ages 8, 10 and 12) and menteal health problems (ie.depression, psychotic experiences, anxiety and self-harm) and sexual relationships.

Outcome, exposure and confounding variables:

Listed as above.

Outcome:

Depression, psychotic experiences and anxiety at 18 years and self-harming behaviour at 16.5 years and sexual relationship in adolescence.

Analysis:

Logistic regression analysis, multiple mediation model and structural equation modelling will be used to test the hypotheses. Given the potential impact of attrition on the findings, the data may be weighted for each part of the analysis. A logistic regression approach to weighting will be employed with bullying data available versus not available specified as the dependent variable and factors likely to predict attrition (e.g., gender, ethnicity, parental social class, family environment) included as independent variables and predicted probabilities (pprob) obtained. The inverse probability weight will then be calculated as 1/pprob.