Weighting for dummies #MRX


Research Team

Few research projects skip the weighting process. It’s an important component that ensures the results are representative of a population. Whether the population is offline census rep, online internet rep, mothers of infants rep, or some other kind of rep, weighting lets us adjust the demographic and psychographic characteristics back to population rates so that the research results will generalize properly to the population.

We could argue forever about the best way to weight. Is it okay to weight data times 3 if the size of that group is 1000? But not ok to weight data times 1.5 if the size of the group is only 25? Where do we draw the line? At what weight and what sample size?

I challenge you, however, to stop weighting your data. Weighting isn’t the solution. Weighting simply gives fewer people a louder voice. If we wanted fewer people, we should have just sampled fewer people.

I challenge you to sample properly. We know the response rates of men and women, young and old, more educated and less educated. We even know the response rates of 21 year old male hispanics with 2 years of college. We know this. So sample for it. If you know you need 50 completes from a low responding group of people, then you’ll have to sample 2000 of them. That’s just how it works.

Sample first. Weight less. Weighting is for dummies.

One response

  1. Wise words, Annie.
    I always look to get the best sampling frame and methods possible from the get go. Yet, even with the best techniques, samples will tend to deviate from the population characteristics. When this happens, THEN weighting the data can help adjust the results to reflect the desired parameters. But weighting in itself won’t “fix” a bad sample…it will only magnify the degree of error in the results.
    There are a lot of techniques that researchers seem to hold out as the “magic pill” for data ills. And for every one of them there is one insurmountable counterpoint:
    why not just take the time to get a better sample in the first place.

    Rob

%d bloggers like this: