© ESO, 2016. This paper presents a new catalog of emission lines based on the GAMA II data for galaxies between 0.07 <z <0.34. The catalog includes four subsamples containing 3000 galaxies drawn from the GAMA II survey and spanning four redshift windows. The four samples are representative of intermediate-mass galaxies down to log M* > 9.4 at z ~ 0.1 and log M* > 10.6 at z ~ 0.30. We have developed a dedicated code called MARVIN that automates the main steps of the data analysis, but imposes visual individual quality control of each measurement. We use this catalog to investigate how the sample selection influences the shape of the stellar mass - metallicity relation. We find that commonly used selection criteria on line detections and by AGN rejection could affect the shape and dispersion of the high-mass end of the M - Z relation. For log M* > 10.6, common selection criteria reject about 65% of the emission-line galaxies. We also find that the relation does not evolve significantly from z = 0.07 to z = 0.34 in the range of stellar mass for which the samples are representative (log M* > 10.6). For lower stellar masses (log M* <10.2) we are able to show that the observed 0.15 dex metallicity decrease in the same redshift range is a consequence of a color bias arising from selecting targets in the r-band. We highlight that this color selection bias affects all samples selected in r-band (e.g., GAMA and SDSS), even those drawn from volume-limited samples. Previously reported evolution of the M - Z relation at various redshifts may need to be revised to evaluate the effect of this selection bias.