Survey Analysis and Missing data in R

About

Representative sample surveys are key to social science whether for
employment numbers or studying attitudes towards the environment. To achieve representativeness we need to consider the sampling process and missing data. Moreover, survey response are usually yes/no, agree to strongly disagree, or categorical (What is your occupation) that require generalized linear models.

The first part of the course discusses survey sampling, and how accounting for sampling weights and strata may lead to more appropriate population estimates. Using the European Social Survey we analyze survey-weighted responses in R.

The second part introduces students to techniques for dealing with missing data - a key thread to representativeness. The focus will be on multiple imputation a flexible technique beyond the survey context.

The third part of course provides a brief introduction to regression models for standard survey responses: binary (yes/no), ordered (Disagree a lot…agree…Agree a lot), and categorical (Occupation/Ethnicity). Followed the application of these generalized linear models accounting for survey sampling and missing data using R.

Course leader

Dr Thees F Spreckelsen

Target group

Students (all groups) and researchers (all groups) in the social sciences (or closely related fields) with an interesting in working with survey data.
Participants should have: Access and introductory experience using R/RStudio and a good
understanding of ordinary least squares regression (OLS).
Self-check: What is an OLS residual and how would you calculate it.

Course aim

Following the course participants will be familiar with key aspects of sample-survey design and analyses methods as well as methods to deal with missing data.
Participants will acquire datamanagement and analysis skills using R to implement these methods.

Fee info