Master's Internship Statistics: Data Analysis and Weighting Methods

Looking for a statistics internship?
Analyze weighting methods and non-response at CBS using R and contribute to better statistics for the Netherlands.

CBS draws samples from the Dutch population to produce various statistics. This involves selective non-response. To correct for this non-response, we perform a so-called weighting, whereby individuals are assigned a weight based on their under- or over-representation in the response. Individuals with background characteristics that are underrepresented in the response are then given a larger weight than individuals with background characteristics that are overrepresented. See, for example, [1] for general information about weighting.

When performing a weighting, there are various factors to take into account. For instance, it may not be possible to correct for skewness for all background characteristics, as the weights would then become too extreme. Additionally, it is often wise not to correct for too many background characteristics at all, because this has a negative influence on the margins of the estimates for the target variables of a study. There is also a relationship with the target variables in the research: at the moment there is a correlation between a background characteristic and the target variables of a study, it is desirable to indeed correct for that background characteristic if that characteristic is skewed in the response. It is therefore often wise to correct for skewness in background characteristics as much as possible.

CBS has a regular procedure for preparing and performing weightings for regular surveys. One option that is not yet used, but seems interesting, is performing a so-called pre-weighting. In this approach, the weighting is performed more or less in two steps, which may allow for correction for more and/or different background characteristics. Perhaps it is possible to first perform a correction for non-response and subsequently perform a further weighting that focuses more on the target variables of the research and the background characteristics that correlate with them. These kinds of questions arise both with the regular samples conducted at CBS and with potential new applications. In this internship, you will work on analyzing different options for the weighting, looking at the differences between these options, and determining which choice is preferred.

Depending on the results, weighting methods could be adjusted, with a direct impact on the statistics of CBS.

[1] https://www.cbs.nl/nl-nl/onze-diensten/methoden/statistische-methoden/throughput/throughput/wegen-als-correctie-voor-non-respons

You have knowledge of and perhaps experience with:

• data analysis;

• various research methods and techniques;

• programming in R;

• reporting findings in a clear and structured form.

Keywords that describe you are proactive, analytical, strong communicator, and critical.

• increase your professional knowledge and gain work experience by conducting research and analyzing data;

• further develop your communication skills and learn to present research results clearly;

• tackle challenges and search for solutions.

In return for your commitment, you will receive an internship allowance of € 800,- gross per month (based on 40 hours per week).

Together with your internship supervisor, you will coordinate how you can work hybridly, with flexible working hours and the option to work partly from home on your internship assignment.

This concerns an internship assignment and not a work experience placement. Before you begin the internship, we expect you to have an internship supervisor from your educational institution, with whom the internship has been coordinated and concrete learning objectives have been agreed upon.
This internship assignment is intended for students who are currently enrolled at an educational institution; graduates are not eligible.