Print

4.2 Data Adjustment

The TwinLife data have been adjusted for some filtering inconsistencies that occur when a respondent answers the filter entry question and the following question(s) in an inconsistent manner with regard to the previous answer. This is mainly a phenomenon in the paper-and-pencil questionnaires (PAPI) that are filled by the respondents without interviewer assistance. In rare cases, there can be programming errors in the CAPI/CASI modules as well which can lead to filter inconsistencies.
The TwinLife data are adjusted for this kind of inconsistency, which means that entries or answers not meeting the filter conditions are deleted. This procedure assumes that the filter entry question was answered correctly whereas the following questions were answered incorrectly.
The constructs that are mainly affected by the adjustment are discrimination (dis), dia (diagnoses), hbe and doc (health-related behavior), spa (academic self-concept), del (delinquent behavior), net (social networks), imo (motivation), sat (domains of life satisfaction), sop (social participation) and mus (cultural capital) because they were at least partly surveyed in a paper-and-pencil questionnaire and included (more or less complex) filter conditions. With the interim data release v4-1-0, which will be provided in autumn 2020, it is planned to release the unadjusted data for all constructs/variables that were at least partly surveyed in the PAPI mode. Users should carefully review whether and which data they use in which way for their analyses. The ➔ TwinLife Technical Report 07 proposes a way how to treat the unadjusted variables that belong to the discrimination construct which is particularly affected by filter inconsistencies.
Please note: The adjustment was not carried out for the igf-variables (intelligence test) in the second face-to-face interview (F2F2), where some of the participants were falsely treated as new members of the sample and took the test a second time. Therefore, the igf-variables of the F2F2 data contain values for participants who should not have filled in the test again. Please consider this when analyzing the igf-variables.