PhD course in Survey Design and Survey Analysis
The course consists of three hours of lectures per day and three hours of computer exercises in Stata. Participants should bring own laptops with Stata installed.
The course falls in two components: Design of a survey and analysis of survey data. It will cover the following topics to the extent time allows:
Design of a survey Discussion of how to make the optimal survey design with respect to obtaining lowest feasible variance of parameter estimates. This part of the course covers choice of strata, relative sample sizes across strata and power calculations to find necessary sample size.
Analysis of survey data Discussion of how to eciently analyze survey data, how to handle stratied data, how to calculate and employ weights, discussion of ex post stratication so that the sample compares to the population on a range of variables, the theoretical background for clustering of standard errors, choice of clustering level, the interaction of weighting and clustering, how clustering aect power calculations, choice of relevant signicance level dependent on sample size and power.