If all other variables are held constant, do small, or large samples produce wider confidence intervals? Why. Small samples, because there is less error. Large samples, because there is more data to include in the interval. Large samples, because z is always larger than t. Small samples, because the sample statistic is less likely to be an accurate predictor of the population parameter