Measuring Information Acquisition from Sensory Input Using Automated Scoring of Natural-Language Descriptions
Figure 2
Error distribution of shared word scores for smaller dataset sizes.
Error of the mean word score, relative to the score for the full normative dataset, as a function of the size of the A) lab-sourced dataset, or B) crowdsourced dataset. There were 100 random samples for each set size. Boxes indicate the interquartile range (IQR), the whiskers contain the range +/− 1.5 * IQR beyond the limits of the boxes (corresponding to 99.7% of the area of a normal distribution).