This page lists answers to frequently asked questions about QuestionLink. If you have further questions, do not hesitate to contact us!
For a detailled description of how the QuestionLink Engine creates the recoding scripts, you can read the following report:
QuestionLink Engine: A methodological and technical Primer on the QuestionLink Engine (PDF) (644 kB)
Latent constructs do not have natural measurement units. Instead, they must be "scaled"; that is anchored to something to become interpretable. In the social scienes results of questions are usually left as raw scores. Thus they are interpetable in terms of the specific question wording and response options. QuestionLink adheres to this convention. Measurement instruments harmonized with QuestionLink thus become interpretable in terms of a chosen reference instrument.
QuestionLink relies on a method called Observed Score Equating in a Random Groups Design (OSE-RG). This harmonization method requires measurements for any two instruments drawn randomly from the same population at roughly the same point in time. This logic is explaned in greater detail in this GESIS Blogpost.
Specifically, QuestionLink relies on the random samples of the adult German population of all the included survey programs. Thus we only process German instruments.
QuestionLink harmonizes the numerical "scales", that is the measurement units of different measurement instruments. In concrete terms, QuestionLink transforms the numerical scores so that if we applied two instruments to the same population, we can expect similarly distributed responses for both instruments. This means that after harmonizing with QuestionLink, we would get similar mean responses, standard deviations, and percentiles for the same population. If the measurement instruments were applied to different populations, we can thus interpret distribution differences as true population differences. This solves many comparability issues.
However, please note that QuestionLink, which uses Observed Score Equating, cannot align random errors. Some measurement instruments tend to be more reliable (i.e., have lower random measurement error) than others. Such differences are not mitigated by QuestionLink. However, please note that this is a common limitation of any harmonization that relies on recoding scores or uses imputation.
QuestionLink harmonizes measurement instruments, not surveys. If other surveys or studies used measurement instruments included in QuestionLink, then you can use the QuestionLink recoding scripts for their data as well.
QuestionLink was always intended to inspire other harmonizaton projects. Especially the automation of Observed Score Equating in a Random Groups Design (OSE-RG) might prove fruitful for other ex-post harmonization projects. Thus we plan in the medium-term to release generalizable components of the QuestionLink Engine. For example, functions to quickly find opportunities to apply OSE-RG based on the instrument-population-time structure of survey programs. In the meantime, feel free to contact us. We will gladly counsel you.