For qualitative and skewed data collected through questionnaires, which do not meet the assumptions of classical regression models, response variables are estimated through alternate regression models. With the objective to obtain significant predictors on the psychological health of young adults, Quantile regression (QR) and Quantile regression Neural Network (QRNN) models were employed on the data collected through three chronological surveys on young adults in higher educational institutions in India during COVID-19 period using the Strength and Difficulty Questionnaire. The QR model was applied with the quantile obtained as ratio of frequencies of two extreme categories.For the QRNN model, optimal quantiles were extracted from a range of quantiles on the basis of the maximum predictive power. The two models were compared through mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE). The efficacy of the models was computed and predictive powers of both the models were obtained under each category of the response variable by comparing the results with the observed data. Significant predictors were extracted through the QR model. Emotional symptom and hyperactivity-inattention problem were found to be the significant predictors in all the three surveys. On the basis of predictive power, QRNN performed better than the QR models. But the significant predictor can’t be extracted through QRNN as the model does not provide the significant value of the test statistic. The study concluded that the two models should be used simultaneously to get a comprehensive picture of data under study.