IX INTERNATIONAL CONFERENCE
CONTEMPORARY ACHIEVEMENTS IN
CIVIL ENGINEERING
BOOK OF ABSTRACTS , 2024.y., pp. 131-131
PARADOXES OF REGRESSION MODELS| |
 |
| DOI: 10.14415/CACE2024.62 |
| UDC: 519.22 |
| CC-BY-SA 4.0 license |
| Author : Trifković, Milan; Kuburić, Miroslav; Nestorović, Žarko |
| |
| | Summary: |
| | Regression analysis is often used to represent the discrete set of points with one line; nevertheless it was straight or curved line which approximates the discrete set of points in plane in “best” possible way. Regression is also suitable to exchange the behaviour of one point in time defined by set of points with one formula. But it also could produce some confusion about obtained results. Namely the exchange of discrete set of points by formula means that coefficient of line should be determined by means of least square method. The further analysis could lead to the insignificance of some coefficients but their elimination from the model could result with significant deviation between real results and improved model. In this research some cases from practice were investigated and discussed. |
| |
| | Keywords: |
| | STATISTICAL HYPOTHESES TESTING, SIGNIFICANCE, ACCURACY |
| |
| |
| |
