Inferential Statistics - the Effect of Computer Graphics Expertise

The images below show results of Kruskal-Wallis statistical significance test calculated on per-pixel basis separately for each scene.

Left column: hover the mouse over the test image to see the reference image.

Right colums: the images show p-values for a particular test, p-values smaller than 0.05 (i.e. statistically significant differences) are visualized in green (when experts average is higher than non-experts) and red (when experts average is lower than non-experts), hover the mouse over the result to see the visualization of p-values smaller than 0.01. The number S shows the overall significance (per scene) and was obtained as the fraction of the number of statistically significant pixels (i.e. p-values<0.05) over the number of pixels in regions marked by at least 5% of observers (i.e. the number of pixels with probabilities of detection >0.05).

apartment - test image	With-ref experiment: Effect of observer experience: S=0, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.031814, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
cg figures - test image	With-ref experiment: Effect of observer experience: S=0.020463, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.093095, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
disney - test image	With-ref experiment: Effect of observer experience: S=0.034548, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.012333, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
kitchen - test image	With-ref experiment: Effect of observer experience: S=0.029649, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.12556, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
red kitchen - test image	With-ref experiment: Effect of observer experience: S=0.030634, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.3432, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
sponza above tree - test image	With-ref experiment: Effect of observer experience: S=0.0017446, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.58744, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
sponza arches - test image	With-ref experiment: Effect of observer experience: S=0.032308, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.21103, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
sponza atrium - test image	With-ref experiment: Effect of observer experience: S=0.061702, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.03733, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
sponza tree shadows - test image	With-ref experiment: Effect of observer experience: S=0.037277, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.10344, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts
sponza trees - test image	With-ref experiment: Effect of observer experience: S=0.0007217, Kruskal-Wallis p < 0.05	With-ref experiment: mean observer data of experts	No-ref experiment: Effect of observer experience: S=0.89481, Kruskal-Wallis p < 0.05	No-ref experiment: mean observer data of experts