Comment on astro-ph/0507588 and astro-ph/0508483

Abstract

Guidorzi has now written two papers (astro-ph/0507588 and astro-ph/0508483, both accepted to MNRAS) on the GRB variability-luminosity correlation in which he finds that expanded samples of L vs. V data are not well described by a power law because the scatter of the data around such a model is more than can be accounted for by the data's statistical errors alone (sample variance) -- "in contrast with the original findings by Reichart et al. (2001)" -- but then proceeds to model these data with a power law anyway and finds significantly shallower L vs. V relationships than Reichart et al. (2001) found. However, as Reichart & Nysewander (2005; astro-ph/0508111) pointed out after Guidorzi's first posting but before his second, Reichart et al. (2001) never modeled their L vs. V data with a power law. Instead, they used a power law with a distribution around it to accommodate and measure this sample variance. Ignoring sample variance in a fit that requires it very easily results in incorrect fitted parameter values due to increased sensitivity to outliers, as well as significantly underestimated uncertainties in these fitted parameter values. Fitting to Guidorzi's own data, Reichart & Nysewander (2005) showed that when sample variance is included in the model, L ~ V^3.4(+0.9,-0.6) with a sample variance of sigma_logV = 0.20(+0.04,-0.04), which is in excellent agreement with the original finding of Reichart et al. (2001) -- L ~ V^3.3(+1.1,-0.9) with a sample variance of sigma_logV = 0.18(+0.07,-0.05) -- when the sample was approximately one-third its current size.