Vol. 8 No. 1 (2025)

Open Access

Original Research Articles

Article ID: 703

Prediction of UCS values using basic geotechnical soil parameters via regression and Artificial Neural Networks ANN

by Mudhaffer Alqudah, Haitham Saleh, Hakan Yasarer, Ahmed Al-Ostaz, Yacoub Najjar

Insight - Statistics, Vol.8, No.1, 2025; 46 Views, 23 PDF Downloads

Unconfined Compressive Strength (UCS) test is a widely used lab procedure for assessing soil’s undrained shear strength. However, conventional lab testing is time-, cost-, and labor-intensive. This study evaluates predictive models for UCS using basic soil parameters. Soil mixtures were prepared and tested through several laboratory experiments, including Atterberg’s limits, particle size distribution, water content, bulk density using Harvard miniature compaction apparatus, and UCS. A total of 152 soil samples were utilized to train the prediction models. To achieve that, multi-linear regression (MLR), multi-nonlinear regression (MNLR), and backpropagation Artificial Neural Networks (ANN) were employed to relate the dependent variable UCS (predicted) to the independent geotechnical parameters (predictors). Results showed that the best model to predict the UCS values for soil using its soil parameters is the ANN-based model with R 2 of 83% and ASE (Averaged Square Error) of 0.0029, followed by the nonlinear regression model with R 2 = 49.2% and ASE of 3.63, and finally the MLR model with R 2 = 44.5% and ASE of 3.92.

Open Access

Original Research Articles

Article ID: 740

Sample selection in social science research: A holistic approach to methodological rigor

by Mohammad Rashed Hasan Polas

Insight - Statistics, Vol.8, No.1, 2025; 37 Views, 18 PDF Downloads

The present study investigates the crucial elements of sample selection in social science research, thoroughly examining the nuances of sampling techniques, categories, and factors. The paper offers a thorough overview of the procedures involved in sampling strategies, with a particular emphasis on non-probability and probability approaches. It also discusses the critical role that sample size determination plays, taking into account variables like cost, ethics, statistical power, accuracy, and generalizability in addition to type I and type II errors. The paper also closely examines how several elements, such as research objectives, design, analytical instruments, and resource constraints, affect the choice of the ideal sample size. The topic of choosing the right data analysis software and how it affects choices about sample size is covered in detail. In the last section of the study, the ideas of power, effect size, and minimum sample size in statistical analysis are thoroughly explored, with a focus on partial least squares structural equation modelling (PLS-SEM).