Новые данные, новая статистика: от кризиса воспроизводимости к новым требованиям к анализу и представлению данных в социальных науках

Социологические исследования. 2018. № 12. С. 30–38.

Deviatko I. F.

The article analyzes main causes and consequences of the interdisciplinary crisis of the reproducibility and reliability of the results of scientific research that has unfolded in the social sciences in parallel with the «data revolution». This crisis is expressed not only in the growing concern of scientists about the reliability of research results and the possibilities to establish the practices securing the transparency of empirical data and the statistical software used for their analysis, but also in disputes on limitations of the routine approach to significance testing and feasibility of alternatives based on Bayesian approach. Some aspects of the relationship between theory and data-driven methods of searching for patterns in empirical data are briefly discussed in the context of describing a new approach to multimodel analysis aiming at evaluation of model robustness and model uncertainty.

Keywords: sociological theory, bayesian approach, publication bias, null hypothesis significance testing, data-driven approach, reproducibility crisis, transparency of data and open data, model robustness, кризис воспроизводимости в социальных науках, открытость данных, публикационное смещение, частотный вывод и байесовский вывод, робастность модели, «ведомый данными» подход и социологическая теория