Monte Carlo simulations are widely used in the social sciences to explore the viability of analytic methods in the face of assumption violations. Simulation results, however, may not be applicable to substantive research applications because they often are conducted under idealized rather than realistic conditions. Shortcomings of simulation design are discussed using linear equations as a case study, focusing on (a) variable distributions, (b) population level specification error, (c) population level measurement precision, and (d) predictor variable relationships. A new strategy is presented, called extant data simulation, which can be used to supplement traditional simulation designs to provide perspectives on Monte Carlo study conclusion generalizability to realistic research scenarios. The approach is illustrated for a binary regression simulation comparing a modified linear probability model to logistic regression. The demonstration results affirm the potential use of a modified linear probability model in a range of analytic contexts, contrary to common recommendations. It also establishes the simplicity and utility of extant data simulation designs for addressing generalizability of traditional Monte Carlo based conclusions. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

中文翻译：

---

使用现有数据模拟总体的蒙特卡洛模拟：修改后的线性概率模型和逻辑回归的应用。

蒙特卡洛模拟在社会科学中被广泛使用，以在面对假设违背的情况下探索分析方法的可行性。但是，仿真结果可能不适用于实质性研究应用程序，因为它们通常是在理想条件下而不是实际条件下进行的。使用线性方程作为案例研究讨论了仿真设计的缺点，重点是（a）变量分布，（b）总体水平规范误差，（c）总体水平测量精度和（d）预测变量之间的关系。提出了一种称为现存数据模拟的新策略，该策略可用于补充传统的模拟设计，以提供有关蒙特卡洛研究结论可推广到现实研究场景的观点。说明了该方法用于将改进的线性概率模型与逻辑回归进行比较的二进制回归仿真。演示结果肯定了在一系列分析环境中修改线性概率模型的潜在用途，这与常见建议相反。它还建立了现有数据仿真设计的简单性和实用性，以解决传统基于蒙特卡洛的结论的普遍性。（PsycInfo数据库记录（c）2021 APA，保留所有权利）。它还建立了现有数据仿真设计的简单性和实用性，以解决传统基于蒙特卡洛的结论的普遍性。（PsycInfo数据库记录（c）2021 APA，保留所有权利）。它还建立了现有数据仿真设计的简单性和实用性，以解决传统基于蒙特卡洛的结论的普遍性。（PsycInfo数据库记录（c）2021 APA，保留所有权利）。