Context.The Square Kilometer Array (SKA) is expected to detect high-redshift galaxies with star formation rates (SFRs) up to two orders of magnitude lower thanHerschelsurveys and will thus boost the ability of radio astronomy to study extragalactic sources. The tight infrared-radio correlation offers the possibility of using radio emission as a dust-unobscured star formation diagnostic. However, the physics governing the link between radio emission and star formation is poorly understood, and recent studies have pointed to differences in the exact calibration required when radio is to be used as a star formation tracer.Aims.We improve the calibration of the relation of the local radio luminosity–SFR and to test whether there are nonlinearities in it.Methods.We used a sample ofHerschelAstrophysical Terahertz Large Area Survey (H-ATLAS) sources and investigated their radio luminosity, which was derived using the NRAO VLA Sky Survey (NVSS) and Faint Images of the Radio Sky at Twenty-cm (FIRST) maps. We stacked the bins of infrared luminosity and SFR and accounted for bins with no detections in the stacked images using survival analysis fitting. This approach was tested using Monte Carlo simulations.Results.After removing sources from the sample that have excess radio emission, which is indicative of nuclear radio activity, we found no deviations from linearity of the mean relations between radio luminosity and either SFR or infrared luminosity.Concluisions.We analyzed the link between radio emission and SFR or infrared luminosity using a local sample of star-forming galaxies without evidence of nuclear radio activity and found no deviations from linearity, although our data are also consistent with the small nonlinearity reported by some recent analyses. The normalizations of these relations are intermediate between those reported by earlier works.