# Details regarding the delete-a-group jackknife

I was reading a paper by Phillip S. Kott on DAGJK:

The delete-a-group jackknife. Journal of Official Statistics, 17 (4):521-526. (full text is freely available)

I don't have much of a survey/sampling background so I'm having a bit of trouble understanding some of the paper.

1. What exactly is a primary sample unit (PSU)? An example would be incredibly helpful.
2. In the second section, Kott assumes that $t=\sum_{h=1}^{H}\sum_{j=1}^{n_h}t_{hj}$ and he defines $q_{hj}=t_{hj}-t_{h+}$, where $t_{h+}=\sum t_{hg}/n_h$ with the summation going over all PSUs in $h$. How did he arrive at the variance formula, $\textrm{Var}(t_{h+})=\frac{n_h}{n_h-1}\sum_{j=1}^{n_h}q_{hj}^2$?
• Here is a good definition of a primary sampling unit. An example could be if I were doing a citywide inspection of street sanitary conditions per street segment in D.C. I'm too lazy to walk around the whole city though. So what I do is take a random sample of a limited number of census tracts, and then take a random sample of street segments within those census tracts. Here the census tract would be the PSU. – Andy W Aug 1 '12 at 23:23
• @AndyW Thanks. That definitely covers my first question. – assumednormal Aug 2 '12 at 0:59