Computing jackknife standard error

[Jarvis Miller]

To confirm that I understand how STATA computes the delete-1 jackknife standard error, I thought of a toy example to replicate. This has turned out to be more difficult than I thought. Jackknife documentation is seen in page 192 http://www.stata.com/manuals13/svy.pdf

 clear use nhanes2jknife.dta svyset, mse jkrweight(jkw_*) vce(jack) svy, mse: mean age height

Is the code used

Now to make sure I understand this:

1. Since probability weights weren't supplied, STATA uses a vector of ones
2. For each variable, STATA uses the jackknife replicate weights to calculate estimates of the mean. So there are R estimates of the mean for age and for height, where R is the number of replicate weights
3. For the R estimates for the mean age, STATA subtracts the original mean (calculated with the weight of ones) from it because mse is specified.
4. For each variable, STATA computes the sum of squares of step 3
5. STATA takes the square root of step 4

This assumes that there are no factors to multiply by in any steps. I'm not sure if this is the case. I wrote code that matches STATAs results (mean and std. err) when you need to compute the jackknife weights when strata, cluster, fpc, etc are provided (which confirms that I'm doing the correct implementation). But the standard error doesnt match when I hand in weights alone and dont provide strata, clsuter, fpc, etc. I think this means I'm missing a factor to multiply by for when weights are provided, but can't find any information about it in the stata documentation. Any help would be appreciated