|
Activity Number:
|
219
- Seeing the World as a Missing Data Problem: Celebrating 40 Years of Multiple Imputation
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, July 30, 2018 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Social Statistics Section
|
|
Abstract #326748
|
|
|
Title:
|
Nonparametric Multiple Imputation for Bridging Between Different Industry Coding Systems
|
|
Author(s):
|
Jörg Drechsler* and Birgit Pech
|
|
Companies:
|
Institute for Employment Research and Amt für Statistik Berlin-Brandenburg
|
|
Keywords:
|
Multiple Imputation;
industry classification;
bridging;
nonparametric;
CART
|
|
Abstract:
|
Industry classifications such as the Statistical Classification of Economic Activities in the European Community (NACE) are regularly updated to ensure that all economic sectors are fully covered. While regular updates are desirable to ensure for example that emerging industries can be classified properly, the changes in the coding system can be a major problem in longitudinal analyses. In most cases a one-to-one mapping between the different versions is not possible which makes a consistent classification for all establishments difficult. In this talk we treat these changes as a missing data problem. The new code is missing for those establishments that only existed while the old code was used and vice versa. We use classification and regression trees (CART) to model the transition probabilities between the classification systems based on years for which both classification systems are available and use these models to impute the missing industry codes. We illustrate that this approach is superior to commonly used strategies such as setting the industry code to the most frequently observed successor/predecessor industry code.
|
Authors who are presenting talks have a * after their name.
