betaselectr: Selective (and Proper) Standardization in Structural Equation Models

Area of Science:

• Psychology
• Statistics
• Data Analysis

Background:

• Standardization enhances result interpretability in psychological research.
• Commonly used structural equation modeling (SEM) programs often standardize results.
• Existing standardization methods can be misleading in specific analytical situations.

Purpose of the Study:

• To address limitations in current standardization practices in SEM and multiple regression.
• To provide accurate standardized coefficients, standard errors, and confidence intervals.
• To develop a tool for researchers to properly handle standardization in complex models.

Main Methods:

• Development of the R package `betaselectr`.
• Implementation of methods to correctly standardize coefficients in three problematic scenarios: dummy variables, moderation product terms, and variables with inherent meaningful units.
• Inclusion of confidence interval calculations that account for sampling error in standard deviations.

Main Results:

• The `betaselectr` package correctly computes standardized coefficients in situations where standard SEM software fails.
• It provides unbiased standard errors and confidence intervals for standardized results.
• The package is applicable to both structural equation modeling and multiple regression.

Conclusions:

• Proper standardization is crucial for accurate interpretation of regression and SEM results.
• `betaselectr` offers a reliable solution for researchers facing common standardization challenges.
• The package improves the quality and trustworthiness of standardized statistical findings in psychological science.