Statistical Inference for the Nested Session Design in Experimental Economics

We identify and describe a class of designs that is very common in experimental economics, to be named 'Nested Session Designs' (NSD's). Analysis and inference of NSD experiments varies greatly, with the nested session effect often ignored. Moreover, since experimental economics datasets exhibit strong departures from both normality and homoskedasticity, the standard ANOVA tools are inadequate for inference. We explore the NSD class and suggest analysis and inference principles based on standard experimental design theory. We also develop nonparametric permutation tests for inference on main effects.