To extrapolate from a sample, the sample must be random. However, all such extrapolations are uncertain, with the uncertainty rising as the sample size approaches zero. No confidence levels are reported, however, so the accuracy of the estimated number of responsive emails (17,853) cannot be discerned.
"Sue the bastards", and "here's to the fourth estate" and all, but could you remind us why this is important? There are no sidebars or old story links, and honestly, it's an outrage a day around here recently.
What constitutes a “preliminary review”?