On the relative threats from sampling and non-sampling errors in print audience