Development of sample designs

Advice on the development of sample designs for face-to-face, written, and telephone-assisted surveys: Our many years of experience help us to develop an appropriate sample plan for every conceivable research problem. We are particularly competent in the field of sampling for international comparative surveys. As we share responsibility for the samples drawn for the European Social Survey, a random sampling system which is unique in Europe.

Calculating design effects

Frequently surveys require complex sampling techniques such as, for instance, multi-stage stratified or clustered sampling. Compared with unlimited random samples, however, these usually show a markedly higher variance. A measure for the latter is the so-called design effect. The design effect is the ratio between the actual variance of a variable under the sampling method used and the variance computed under the assumption of simple random sampling. GESIS supports users in estimating design effects. These play a particular role in the planning of survey size.

Weighting

Frequently surveys use sample plans with unequal probabilities of selection. For instance, in telephone surveys one target person is usually selected within the household contacted. Here, the probability of inclusion varies depending on the number of household members who meet the criteria of the population. These diverging probabilities need to be taken into consideration in the later process of data analysis by means of weighting. Aside from design weighting, in many cases an adjustment weighting is carried out which is intended to adjust sample biases caused by nonresponse to known distributions of the population, for instance with regard to age and sex. GESIS provides support with the calculation of weighting factors for both design and adjustment weightings.

Telephone sampling service

Alongside our support in issues relating to sampling we offer our users a telephone sampling service from a selection frame specially developed by us. (only in German)