Greedy knot selection algorithm for restricted cubic spline regression
Non-linear regression modeling is common in epidemiology for prediction purposes or estimating relationships between predictor and response variables. Restricted cubic spline (RCS) regression is one such method, for example, highly relevant to Cox proportional hazard regression model analysis. RCS r...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2023-12-01
|
| Series: | Frontiers in Epidemiology |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fepid.2023.1283705/full |
| _version_ | 1797387077212438528 |
|---|---|
| author | Jo Inge Arnes Alexander Hapfelmeier Alexander Horsch Tonje Braaten |
| author_facet | Jo Inge Arnes Alexander Hapfelmeier Alexander Horsch Tonje Braaten |
| author_sort | Jo Inge Arnes |
| collection | DOAJ |
| description | Non-linear regression modeling is common in epidemiology for prediction purposes or estimating relationships between predictor and response variables. Restricted cubic spline (RCS) regression is one such method, for example, highly relevant to Cox proportional hazard regression model analysis. RCS regression uses third-order polynomials joined at knot points to model non-linear relationships. The standard approach is to place knots by a regular sequence of quantiles between the outer boundaries. A regression curve can easily be fitted to the sample using a relatively high number of knots. The problem is then overfitting, where a regression model has a good fit to the given sample but does not generalize well to other samples. A low knot count is thus preferred. However, the standard knot selection process can lead to underperformance in the sparser regions of the predictor variable, especially when using a low number of knots. It can also lead to overfitting in the denser regions. We present a simple greedy search algorithm using a backward method for knot selection that shows reduced prediction error and Bayesian information criterion scores compared to the standard knot selection process in simulation experiments. We have implemented the algorithm as part of an open-source R-package, knutar. |
| first_indexed | 2024-03-08T22:19:35Z |
| format | Article |
| id | doaj.art-b0f8f426661f4162ab9796e2ab17c41d |
| institution | Directory Open Access Journal |
| issn | 2674-1199 |
| language | English |
| last_indexed | 2024-03-08T22:19:35Z |
| publishDate | 2023-12-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Epidemiology |
| spelling | doaj.art-b0f8f426661f4162ab9796e2ab17c41d2023-12-18T15:57:18ZengFrontiers Media S.A.Frontiers in Epidemiology2674-11992023-12-01310.3389/fepid.2023.12837051283705Greedy knot selection algorithm for restricted cubic spline regressionJo Inge Arnes0Alexander Hapfelmeier1Alexander Horsch2Tonje Braaten3Department of Computer Science, Faculty of Science and Technology, UiT The Arctic University of Norway, Tromsø, NorwayInstitute of AI and Informatics in Medicine, TUM School of Medicine, Technical University of Munich, Munich, GermanyDepartment of Computer Science, Faculty of Science and Technology, UiT The Arctic University of Norway, Tromsø, NorwayDepartment of Community Medicine, Faculty of Health Sciences, UiT The Arctic University of Norway, Tromsø, NorwayNon-linear regression modeling is common in epidemiology for prediction purposes or estimating relationships between predictor and response variables. Restricted cubic spline (RCS) regression is one such method, for example, highly relevant to Cox proportional hazard regression model analysis. RCS regression uses third-order polynomials joined at knot points to model non-linear relationships. The standard approach is to place knots by a regular sequence of quantiles between the outer boundaries. A regression curve can easily be fitted to the sample using a relatively high number of knots. The problem is then overfitting, where a regression model has a good fit to the given sample but does not generalize well to other samples. A low knot count is thus preferred. However, the standard knot selection process can lead to underperformance in the sparser regions of the predictor variable, especially when using a low number of knots. It can also lead to overfitting in the denser regions. We present a simple greedy search algorithm using a backward method for knot selection that shows reduced prediction error and Bayesian information criterion scores compared to the standard knot selection process in simulation experiments. We have implemented the algorithm as part of an open-source R-package, knutar.https://www.frontiersin.org/articles/10.3389/fepid.2023.1283705/fullmodel selectionnon-linear regressionpredictionrestricted cubic splinesalgorithm |
| spellingShingle | Jo Inge Arnes Alexander Hapfelmeier Alexander Horsch Tonje Braaten Greedy knot selection algorithm for restricted cubic spline regression Frontiers in Epidemiology model selection non-linear regression prediction restricted cubic splines algorithm |
| title | Greedy knot selection algorithm for restricted cubic spline regression |
| title_full | Greedy knot selection algorithm for restricted cubic spline regression |
| title_fullStr | Greedy knot selection algorithm for restricted cubic spline regression |
| title_full_unstemmed | Greedy knot selection algorithm for restricted cubic spline regression |
| title_short | Greedy knot selection algorithm for restricted cubic spline regression |
| title_sort | greedy knot selection algorithm for restricted cubic spline regression |
| topic | model selection non-linear regression prediction restricted cubic splines algorithm |
| url | https://www.frontiersin.org/articles/10.3389/fepid.2023.1283705/full |
| work_keys_str_mv | AT joingearnes greedyknotselectionalgorithmforrestrictedcubicsplineregression AT alexanderhapfelmeier greedyknotselectionalgorithmforrestrictedcubicsplineregression AT alexanderhorsch greedyknotselectionalgorithmforrestrictedcubicsplineregression AT tonjebraaten greedyknotselectionalgorithmforrestrictedcubicsplineregression |