paper.md

In padpadpadpad/nlsTools: Methods for Finding the Best Estimated Fits for Non-Least Squares Regression over the Level of a Factor

title: 'nlsLoop' tags: - non-linear regression - R authors: - name: Daniel Padfield orcid: 0000-0000-0000-1234 affiliation: "1" affiliations: - name: University of Exeter index: 1 date: 14 February 2016 bibliography: paper.bib

Summary

• nlsLoop is a simple R package that gives a more reproducible and reliable method for fitting individual non-linear regression fits over levels of a factor. This procedure is commonly done using nlme::nlsList, but this function only uses one set of parameter values. Consequently, some curves fail to converge on the correct parameter values simply because of the starting values being too far away from the starting values. nlsLoop improves on nlme::nlsList by allowing multiple starting values for each parameter, thereby exploring more parameter space when model fitting. The best model is chosen based on AIC score, ensuring that results are more reproducible and replicable.
• This method and the R package nlsLoop have been used in multiple publications in the last two years (Padfield et al. 2016, Padfield et al. 2017). It is also used in other projects within the lab group and is likely to be useful to the wider scientific community.

References

• Padfield, D., Yvon-Durocher, G., Buckling, A., Jennings, S. & Yvon-Durocher, G. (2016). Rapid evolution of metabolic traits explains thermal adaptation in phytoplankton. Ecology Letters, 19, 133–142. doi:10.1111/ele.12545
• Padfield, D., Lowe, C., Buckling, A., Ffrench-Constant, R., Jennings, S., Shelley, F., et al. (2017). Metabolic compensation constrains the temperature dependence of gross primary production. Ecology Letters, 20, 1250–1260. doi:10.1111/ele.12820

padpadpadpad/nlsTools documentation built on May 24, 2019, 5:59 p.m.