marginally significantp < .08
Using these criteria, we found that sampling criteria using the SNS, the CRT, and SAT Math questions produced marginally significant ( p < .08) improvements in subsample-wide forecasting accuracy without significant change in confidence.