Synthetic data effectively balanced low-incidence groups, while maintaining analytical conclusions closely aligned with the original dataset.
TL;DR
Synthetic data effectively replicated utility distributions, with average differences remaining within single digits even at 25% synthetic substitution.
What did we find out?
MaxDiff analysis, a common method for preference modeling, was tested using synthetic data from 414 respondents in a new US only survey. Since men under 35 are typically harder to reach, we simulated data for this group. While market researchers recognize the challenge of small or imbalanced samples, they often accept these limitations due to time and budget constraints. Synthetic data, however, offers a way a way to bridge this gap.
How did we do it?
Assessing the value of increasing niche sample sizes in advanced statistical estimation.
What was the experiment?
Enhancing MaxDiff analysis with synthetic data
EXPERIMENT #3
