“…We stress that a large number of discordancy tests for univariate data (Dixon, 1950(Dixon, , 1953Grubbs, 1950Grubbs, , 1969Grubbs and Beck, 1972;Barnett and Lewis, 1994;Verma, 1997Verma, , 2005Verma, , 2012Verma and Quiroz-Ruiz, 2008;Verma et al, 2008Verma et al, , 2014Verma and Díaz-González, 2012) were applied to the complete dataset to assure that the remaining discordant outlier-free data were normally distributed. This procedure has been successfully used by us in numerous other studies (e.g., Verma, 2013;Armstrong-Altrin, 2013, 2015;Verma et al, 2014Verma et al, , 2015a. Distributions other than the Gaussian are not likely for such experimental data, especially when log-ratios are involved (e.g., Aitchison, 1986;Law and Kelton, 2000;Verma, 2005Verma, , 2015a.…”