Errors in Logic and Statistics Plague a Meta-Analysis (response to Andow and Lövei 2012): Table 1.

Abstract

As we noted previously (Shelton et al. 2009a, b), we strongly believe in the power of meta-analyses to help advance our collective understanding of the potential risks of Bt crops for nontarget organisms by identifying negative, neutral, and positive effects of the technology in both laboratory and field studies. Although we agree on this point, it is equally important that such studies do not contain errors in logic or statistics. We acknowledge that Andow and Lovei (2012) have corrected a statistical error in their previous publications (Lovei et al. 2009), but point out another statistical error (see below) in their latest letter. However, more important than these statistical errors, we question their continued attribution of hazard to a protein rather than, more accurately, to poor prey or host quality. We believe this is an error in logic. Therefore, we strongly oppose the latest statement by Andow and Lovei (2012) that their conclusions of detectable nonzero effects of Cry proteins on nontarget organisms were “… criticized by Shelton et al. (2009a, b) on statistical grounds.” They missed the point again. Our primary criticism then and now is that they continue to ignore prey and host-quality effects and the ecological context in their effort to inform risk assessment. Besides the fault in logic of ignoring prey or host quality, we also take issue with the statement in their latest letter (Andow and Lovei 2012): “We conclude not that there could have been type II error, but that there has been type II error.” Given the nature of …