Solutions
- Print Media Research
- Discover how your magazine’s readers believe your publication can be improved.
- Online panels
- Feedback from customers on a weekly basis? We have weekly contact with thousands of online participants, allowing us to gather together millions of variables to help you adapt to customer’s needs, instantly.
- Mixxit
- Gain insight into the consumption behavior of Dutch citizens of Turkish, Moroccan, Surinam or Antillean descent with the aid of our online panel, made up of over 7,500 15-40 year olds.
Reliability
In order to be assured of qualitative, reliable and representative research, Motivaction has in-house methodologists to monitor research projects. An important component in research projects is the sampling procedure.
Propensity Sampling
In an Internet panel in which people participate on a voluntary basis, certain groups of people are overrepresented, while others are underrepresented. For example, online panels tend to contain a relatively high number of young and highly-educated people. However, with respect to other variables, panels deviate from the population.
If you would want to carry out a representative sample survey, you must take these deviations into account. To achieve this goal, Motivaction uses a technique called propensity sampling. This advanced technique makes it possible to take random samples, which provide an optimal reflection of the population with regard to a large number of aspects.
Propensity sampling basically means that lots are drawn for every respondent to determine whether or not he or she will be included in the random sample. The chance of being chosen for the random sample varies from one respondent to the next. This probability is determined on the basis of the respondent's background characteristics, including age, education and norms and values patterns. For example, if there are relatively few respondents in a panel with a low education, those with a low education will have a better chance of ending up in the random sample than those with a higher education.
Random samples in online panels - a new approach (PDF)
Weighting
If the random sample is not representative in other words, if it displays systematic deviations from the target population then these deviations may be corrected. This is referred to as re-weighting the random sample. This means that we will assign a higher weight to some of the respondents (weighting factor > 1) than others (weighting factor < 1). In most cases, however, we use a Rim-weighting or propensity weighting method.
Rim Weighting
'Rim' stands for a margins. In a Rim weighting, we equate the margins from the sample to those of the target population. This occurs in an iterative process: we continue to calculate until the conditions are satisfied.
Propensity Weighting
It is not always possible to correct for margins. In the case of a panel study, for example, there are no non-users of the Internet participating, even though we would like to make statements about this group. It is therefore not possible to assign a higher weighting factor to the non-users of Internet since they are not included in the panel. Motivaction corrects these types of random samples for population statistics using propensity weighting.
Literature about propensity:
Joffe, M.M. and Rosenbaum, P.R. (1999), Propensity scores. 'American Journal of Epidemiology'. 150, 4, August 15, 1999.
Little, R.J., and Rubin, D.R., (1999), Comment. 'Journal of the American Statistical Association: Theory and Methods'. December 1999, vol. 94, 448, pp. 1130-1132.
Rubin, D.R., and Thomas, N. (2000), Combining propensity score matching with additional adjustments for prognostic covariates. 'Journal of the American Statistical Association: Theory and methods'. June 2000, vol. 95, 450, pp. 573-.