Recently , Guidorzi et al .
( 2005 ) expanded the size of the sample of GRBs for which variabilities and peak luminosities have been measured , from 11 to 32 .
They confirm the existence of a correlation , but find a dramatically different relationship between L and V than had originally been found .
We find that this is the result of improper statistical methodology .
When we fit a model to the data that accommodates both statistical variance ( in two dimensions ) and sample variance , we find that L \sim V ^ { 3.4 ^ { +0.9 } _ { -0.6 } } with a sample variance of \sigma _ { \log { V } } = 0.20 ^ { +0.04 } _ { -0.04 } , which is consistent with the original finding of Reichart et al .
( 2001 ) – L \sim V ^ { 3.3 ^ { +1.1 } _ { -0.9 } } with a sample variance of \sigma _ { \log { V } } = 0.18 ^ { +0.07 } _ { -0.05 } – and inconsistent with the finding of Guidorzi et al .
( 2005 ) : L \propto V ^ { 1.3 ^ { +0.8 } _ { -0.4 } } with sample variance assumed to be zero .