There’s a variety of differing weighting methods to weight your data accurately, such as: rim weighting, sample weighting, ranking or iterative proportional fitting, propensity weighting and cell weighting. How do you choose which method you’ll use to get a balanced and representative sample?
Before you make a choice you’ll want to remember that all these different methods pretty much do the same thing. There are minor differences in both execution and outcome when using these techniques, as the outcomes are all relatively the same.
The most important thing to take into account when choosing between weighting methods is the variables that you’re using to control the sample. Researchers tend to use the ‘standard’ demographics such as age, gender, income and household size. If you want to get the best results however, you’ll think long and carefully about what demographic variables have a serious and actual impact on the dependent variable you’re studying for your client.
Let’s say you’re studying attitudes in a population or certain behaviours regarding a product or a service. The most critical variable isn’t the age or household size of the participant, but whether or not they’ve actually experienced the product, if they’ve bought it or have come into contact with it. If you want to do your research well, you’ll want to control variables like these – which are actually relevant to your weight outcomes, unlike their age or their gender.
When you want to be certain that you’re sample is representative for the population you’ve studied you can use a technique or procedure called statistical weighting. If you’re looking for a representative sample, it has to be of the same composition as the population that you’re studying.