We'll conclude module 1 of course 5 by talking about the effects of weighting on standard errors. Now, there are a number of features of a design that can affect standard errors, and I've listed them here. The ones I have in mind are stratification, clustering, and sampling with varying probabilities. So let's think about stratification first. An efficient allocation to the stratum can reduce standard errors, at least for full population estimates. Another use of standard errors is to control the sample size and precision of stratum estimates. So we need to think about both of those when we estimate standard errors. Clustering is the second thing that I mentioned that will affect standard errors. Oftentimes, that will increase standard errors. It doesn't have to be that way, but if you cluster units together that are somewhat alike on what you're measuring, that will be the effect. You'll get less precision from your sample size than if the units were completely independent and unclustered. The effects can be different for full population estimates as compared to domain estimates or subgroup estimates. The effect can be different on different y values. And it can be different for different statistics, totals, means, model parameter estimates. All those things may be affected differently by stratification, and clustering, and the use of weights. Now finally, weighting adjustments can increase or decrease standard errors. And depending on exactly what the purpose of the weighting adjustments are, non-response estimates will tend to increase standard errors, at least in typical applications. If you use calibration to population controls, those can decrease standard errors, assuming that you've got covariates used in the calibration that are reasonably good predictors of the y's that you're trying to measure. So that module is an overview of the purposes of weighting. In the next module, we'll cover specific steps that need to be used.