Teribus:

Yes exactly Guest TIA, it is a method used by pollsters, "Batch Sampling" does not and cannot EVER give you anything other than the roughest of estimations - It cannot in any way be treated as any sort of serious study or attempt to quantify civilian casualties.

Except that it is same method used for casualty figures in the Congo and Darfur, which are accepted because they don't have the political implications of the Iraq numbers (see my link to discussions above).

This garbage about "the roughest of estimates" is just that. Statistical sampling is an "estimate" (in fact, that's what it's designed to be; if you wanted the actual number, you'd count everything and wouldn't sample ... be my guest, Teribus, go over and count 'em all). You can use various methods to reduce the uncertainty (narrow the "confidence interval"). Increasing sample size does that. If you do totally random sampling, you reduce it the most, but if that's not practical, you can reduce it by counting a "cluster" (as they did the the study) of houses physically adjacent.   This reduces variability, but you can account for that as well by looking at "within-group variance" and "between-groups variance" to estimate by how much the variance is reduced (and thus the effective sample size). This they did. It's a perfectly legitimate methodology, used in situations where the costs (and risks) of completely random sampling are too high (such as in war zones where your researchers are in danger of getting killed just to get your data). And it increases the sample size over just looking at the first home selected (albeit not as much as selecting the same number of houses completely at random).

IOW, you're just ignerrent of statistics ... and science, for that matter.

Cheers,