Abstract:School bullying victimization is a variable that cannot be measured directly. Taking into account that this variable has a lower bound, given by the absence of bullying victimization, this paper proposes IRT logistic models, where the latent parameter ranges from $0$ to $\infty$ or from $0$ to a positive real number R, defining the IRT parameters and proposing an empirical anchor procedure. As the academic abilities and the school bullying victimization can be explained due to associated factors such as habits, sex, socioeconomic level and education level of parents, IRT regression models are proposed to make joint inferences about individual and school characteristic effects. Results from the application of the proposed models to the Bogotá school bullying dataset are presented. The need for testing based in statistical models increases in different fields.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to measure and model the variable of school bullying victimization, which cannot be directly measured, more accurately. Specifically, the author proposes new item response theory (IRT) models to adapt to the characteristics of school bullying victimization, and analyzes the Bogotá school bullying data through these models.
### Main problems:
1. **Measuring non - directly observable variables**: School bullying victimization is a latent variable that cannot be directly measured. Traditional IRT models assume that the range of latent variables is from negative infinity to positive infinity, which may be inappropriate in practical applications because bullying victimization has a clear lower limit (i.e., no bullying behavior).
2. **Developing IRT models suitable for specific contexts**: In order to better adapt to the actual situation of school bullying victimization, the author proposes some new IRT models, in which the range of latent variables is from 0 to positive infinity or from 0 to a certain positive real number \( R \). These models can more accurately reflect the degree distribution of bullying victimization.
3. **Jointly inferring the influence of individual and school characteristics**: Since academic ability and school bullying victimization can be explained by some related factors, such as habits, gender, socioeconomic level and parental education level, an IRT regression model is proposed in order to simultaneously infer the influence of individual and school characteristics.
### Specific methods:
- **New IRT models**: Two new IRT models are proposed. One is a model with an ability parameter range of \( (0, \infty) \), and the other is a model with an ability parameter range of \( (0, R) \). These two models use logarithmic transformation and logit, probit or log - log transformation to process the ability parameter respectively.
- **Bayesian estimation**: The Bayesian method is used to estimate the model parameters, and Gibbs sampling is used for posterior inference.
- **Anchoring procedure**: An empirical anchoring procedure is introduced to determine the performance levels and associate these levels with specific questionnaire items, thereby giving substantial interpretations to the measurement results.
### Application:
The paper applies the proposed models to the Bogotá school bullying data set, verifies the validity of the models, and shows the performance of different models in describing the distribution of bullying victimization. The results show that the model with a bounded range and logit transformation (Model 3a) is the most suitable for describing bullying victimization.
### Summary:
This paper aims to more accurately measure and understand school bullying victimization by improving IRT models, and provide a scientific basis for formulating effective prevention policies.