Effect of each parameter on z scores, as estimated by model averaging of the top four models in the set (those with an Akaike weight > 0.05).

Note that the predictor and response variables were rescaled to have a mean of 0 and standard deviation of 0.5 prior to running the models [37]. The FoR category × Blindness interaction was not among the top models, indicating it had no detectable effect on z score. The effects of each FoR category are shown relative to the reference level “Agricultural and veterinary sciences.” The Importance column gives the summed Akaike weights of all the models containing the focal predictor; high importance shows that a parameter was present in many of the best models, and was thus a good predictor of z scores relative to the other measured variables.