Imputation Short Courses
We will be offering two half-courses dealing with imputation and missing data. A UCT certificate of attendance will be given to participants who attend both courses.
1. Dealing with missing data in household surveys, 22-24 November 2023
Missing data are a serious issue for the analysis of household data. In this course, we will explore approaches for dealing with this matter. The course is designed to:
• Highlight why missing data are important,
• Provide an overview of different methods for dealing with missing data, including weighting, different types of simple imputations, as well as multiple imputation,
• Look at the implication of different approaches for the estimates of means, proportions and percentiles, as well as the accuracy of those estimates,
• Show what can go wrong when missing data are handled badly.
Provisional Schedule:
Day 1: Missing data in social surveys. Types of missingness. Example: Earnings Data. Weighting approaches
Day 2: Imputation approaches. Simple imputations: parametric and hotdeck. Problems with imputation. Accuracy of statistics using imputation.
Day 3: Introduction to multiple imputation. Patterns of missingness. Practical issues. What is the imputation for?
Prerequisites: Participants are expected to have a good grounding in multivariate analysis techniques (multiple regression, discrete choice models), handling of survey data and some experience with Stata.
2. Multiple Imputation using Stata, 27-29 November 2023
This course examines the tools available in Stata to run multivariate estimation routines using multiple imputation. This course is designed to:
• Introduce the Stata suite of multiple imputation tools,
• Provide hands-on experience in analysing multiply imputed data, particularly in labour economics,
• Show what can go wrong when imputation is done badly.
Provisional Schedule:
Day 1: Univariate imputations if the pattern is monotone. Data handling, transformation, and creation with multiply imputed data. Importing and exporting. Descriptive statistics.
Day 2: Regression methods with multiply imputed data. Diagnostics. More complex methods of imputation. Chained imputations. Multivariate imputations.
Day 3: Pitfalls with imputations. Imputations and measurement errors. Alternatives. Documenting the process.
Prerequisites: Completion of the “Dealing with missing data in household surveys” or equivalent.
Mode of Delivery: We hope to run the courses in person at the University of Cape Town but we may offer it online if we don’t have enough people attending in person. Sessions will run from 9.00am to 4.00pm. There will be a strong practical component, using Stata.
Instructor: Professor Martin Wittenberg, School of Economics and Director of DataFirst
Course Fees: Each course costs R7000.00. If you choose to register for both courses, the total fee would be R12 500.00. Partial scholarships are available to bona fide students
Closing Date: These courses have been cancelled.
*DataFirst reserves the right to cancel if there is not enough interest in the courses.
