Statistical processing

Source data

The gross sample size of the AES survey was 7.868 respondents. The interviews was conducted by a combination of web-interviews and interviews by phone.

Different administrative sources are used to select the sample. Administrative sources are also used to obtain various background information on the people interviewed, for example on educational level or workplace.

Population: The Danish population of age 25-64 years.

Sampling Frame: Central Person Register (CPR).

Frequency of data collection

Approx. every 5-6th year.

Data collection

The persons selected were checked for telephone number, and contacted by phone if possible. The persons for whom a phone number could not be identified were contacted by letter asking either to get in touch with Statistics Denmark by phone or alternatively to answer the interview on the web. Web-interviews are used as a supplement to telephone interviews, because the experience from other surveys has shown it is some certain age groups who can't be reached by phone. In other words the usage of web-interviews increases both the representativeness and the as well as the response rate.

Data validation

In order to assure a high quality of the data some control procedures were build-in in the as well in the web as in the telephone set-up for the interviewing. Data was also edited for validity and consistency between the answers.

Data compilation

A weighting procedure on sex and age was carried out in order for the sample to reflect the distribution of the two variables in the universe. For each of the three age groups the results were enumerated to represent the total universe.

The weighting procedure in the AES is based on a regression estimator. In the regression estimator auxiliary information is exploited in the estimation procedure. In the case of the Danish AES survey the auxiliary information comes from the statistical registers of Statistics Denmark. The basic idea in the regression estimator is that relevant auxiliary information (chosen in a way that it covariates with variables of interest in the survey) contain information about study variables. This information is utilized in the estimation process to obtain better estimates. The stronger the correlation between target variables and auxiliary variables, the better is the final estimate. The quality is improved by lowering the standard error on estimates and/or by correcting for non response bias. An important characteristic of the regression estimator is the calibration property. Since information from administrative registers is used as auxiliary information the final estimates are based on distributions from administrative registers and not from the response population. This means that the final estimates are calibrated according to the auxiliary information from administrative registers. The estimation is carried out using the SAS-macros CLAN developed by Statistiska cenrtralbyrån in Sweden. The variance (and standard error) estimate is based on a Taylor approximation.

Adjustment

No further corrections are carried out apart from what has been described in Data Validation and Data Treatment.