Predictors of Sexual Harassment Using Classification and Regression Tree Analyses and Hurdle Models: A Direct Replication

Jan Louw Kotzé, Patricia A. Frazier, Kayla A. Huber, Katherine A. Lust

Research output: Contribution to journalArticlepeer-review


Sexual harassment affects a large percentage of higher education students in the US. A previous study identified several risk factors for sexual harassment using hurdle models and classification and regression tree (CART) analyses. The purpose of the present study was to assess the robustness of these findings by replicating the analyses with a new sample of students. Secondary data analysis was conducted using data from 9,552 students from two- and four-year colleges. Hurdle model coefficients were assessed for replicability based on statistical significance and consistency of the replication effect size relative to the original effect size. Kotzé et al.’s findings were robust, with 91% of all tested effects meeting at least one of two replication criteria in the hurdle models and 88% of the variables replicating in the CARTs. Being younger, consuming alcohol more frequently, attending a four-year college, and having experienced more prior victimization and adversity were important predictors of peer harassment whereas being LGBQ+ was an important predictor of sexual harassment from faculty/staff. These findings can inform targeted prevention and intervention programs. More research is needed to understand why certain demographic and contextual variables are associated with greater harassment risk.

Original languageEnglish (US)
JournalJournal of Sex Research
StateAccepted/In press - 2023

Bibliographical note

Publisher Copyright:
© 2023 The Society for the Scientific Study of Sexuality.

PubMed: MeSH publication types

  • Journal Article


Dive into the research topics of 'Predictors of Sexual Harassment Using Classification and Regression Tree Analyses and Hurdle Models: A Direct Replication'. Together they form a unique fingerprint.

Cite this