Non-Gaussian error distribution of 7Li abundance measurements

We construct the error distribution of Li-7 abundance measurements for 66 observations (with error bars) used by Spite et al. (2012) that give A(Li) = 2.21 +/- 0.065 (median and 1 sigma symmetrized error). This error distribution is somewhat non-Gaussian, with larger probability in the tails than is predicted by a Gaussian distribution. The 95.4% confidence limits are 3.0 sigma in terms of the quoted errors. We fit the data to four commonly used distributions: Gaussian, Cauchy, Student's t and double exponential with the center of the distribution found with both weighted mean and median statistics. It is reasonably well described by a widened n = 8 Student's t distribution. Assuming Gaussianity, the observed A(Li) is 6.5 sigma away from that expected from standard Big Bang Nucleo-synthesis (BBN) given the Planck observations. Accounting for the non-Gaussianity of the observed A(Li) error distribution reduces the discrepancy to 4.9 sigma, which is still significant.