Author

Kendall Moore

Date of Award

Summer 2019

Document Type

Open Access Dissertation

Department

Psychology

First Advisor

Amanda Fairchild

Abstract

Given recent calls for advancing valid instrumentation in the field of cyberaggression, the present study evaluated construct validity and measurement invariance for the Cyberbullying Experiences Survey (CES) in a high school and college student sample. A series of confirmatory factor analyses (CFA), reliability analyses, and a nomological net evaluation were conducted to address these aims. The data did not provide support for the hypothesized four-factor model for cyberaggression or cybervictimization (i.e., unwanted contact, malice, deception, and public humiliation). Upon implementing suggested and theoretically supported modification indices, support for a four-factor solution for both cyberaggression and cybervictimization was provided.

To subsequently evaluate measurement invariance, single-group CFAs were constructed to test invariance of the four-factor structure across college and high school students. Results provided support for the four-factor model solution of cyberaggression and cybervictimization in the college sample but not in the high school sample. Two cyberaggression subscales (i.e., unwanted contact and deception) correlated at r = .99, indicating the potential for multicollinearity, and incremental fit indices for the cybervictimization model solution did not meet recommended cut-off values in the high school sample. Revised model results based on statistical and theoretical considerations evaluated a restructured three-factor solution for cyberaggression (i.e., “sexual,” “direct,” and “coercion”) and cybervictimization (i.e., “sexual,” “direct,” and “defamation”). Fit indices provided initial support for the revised model solution for both CES cyberaggression items (College: MLM c2 (163) = 273.01, RMSEA = .04, CFI = .92, SRMR = .06; High School: MLM c2 (165) = 196.29, RMSEA = .03, CFI = .96, SRMR = .08) and cybervictimization items (College: MLM c2 (163) = 367.81, RMSEA = .05, CFI = .93, SRMR = .06; High School: MLM c2 (160) = 256.32, RMSEA = .06, CFI = .92, SRMR = .07).

Utilizing the revised factor solution for the remaining analyses, the CES displayed evidence for internal consistency reliability across college (cyberaggression items: α = .83; cybervictimization items: α = .89) and high school (cyberaggression items: α = .88; cybervictimization items: α = .90), although internal consistencies for the CES cyberaggression subscales ranged from poor to good (α = .54 - .88) and acceptable to excellent (α = .76 - .92) for the CES cybervictimization subscales across both college and high school samples. Evidence for convergent validity with theoretically similar constructs was mixed. Specific areas of model misspecification as well as directions for future cyberaggression measurement research and policy are discussed.

Share

COinS