Analysis of survey results using the stepwise method in the model of logistic regression on the platform of MS Excel (the survey of Ukrainian students in February 2024)

Abstract

There are presented base formulas of logistic regression and as exactly they can be realized on the platform of MS Excel (Real Statistics, Logistic of and of Probit Regression). The used statistical measures of the productivity of tests of binary classification are considered widely: Sensitivity, Specificity, Positive Predictive Value (PPV), Negative Predictive Value (NPV), prevalence of case (Prevalence) and other. The content of McFadden's, Cox and Snell's and Nagelkerke's coefficients of determination is considered. Also, we give the detailed analysis of curve of Receiver Operating Characteristic – ROC curve analysis, as well as application of Youden's J statistic.

The model was evaluated using performance indicators such as sensitivity (71.4%), specificity (92%), and the area under the ROC curve (AUC = 0.81), indicating high model accuracy. Additionally, the odds ratio for key factors showed that an increase in the impact of war-related stress by 1 unit decreases the probability of continuing education by 28.57%. The results obtained provide a better understanding of how external factors affect students' psychological state and academic performance, allowing for the development of recommendations to improve student support in crisis conditions.