Monte Carlo simulations using extant data to mimic populations: Applications to the modified linear probability model...

Area of Science:

• Social Sciences
• Statistical Modeling
• Computational Statistics

Background:

• Monte Carlo simulations are crucial in social sciences for evaluating analytical methods under assumption violations.
• Traditional simulations often use idealized conditions, limiting the applicability of their findings to real-world research.
• Existing simulation designs may not fully address the complexities of realistic research scenarios.

Purpose of the Study:

• To discuss shortcomings in traditional Monte Carlo simulation designs within the social sciences.
• To introduce and illustrate a new strategy, extant data simulation, for enhancing the generalizability of simulation results.
• To compare a modified linear probability model with logistic regression using extant data simulation for binary regression.

Main Methods:

• Analysis of Monte Carlo simulation design limitations, including variable distributions, specification error, measurement precision, and predictor relationships.
• Development and application of the extant data simulation strategy.
• Binary regression simulation comparing a modified linear probability model against logistic regression.

Main Results:

• Extant data simulation provides a valuable supplement to traditional simulation designs, improving generalizability.
• Demonstration results support the utility of a modified linear probability model in various analytic contexts, challenging common recommendations.
• The study confirms the simplicity and effectiveness of extant data simulation for assessing the generalizability of Monte Carlo conclusions.

Conclusions:

• Extant data simulation enhances the relevance of Monte Carlo simulation findings for substantive social science research.
• The modified linear probability model shows potential utility in binary regression, contrary to some existing recommendations.
• This approach offers a practical method to bridge the gap between simulation studies and realistic research applications.