By “noise function”, I was referring to the distribution you are wrapping around your predicted mean. You’re predicting that RTs should be:
mu = pm.math.dot(inter, alpha) + pm.math.dot(slope, beta)*Worry
but then you wrap this in a Student t or a normal or an an ExGaussian. I, personally, would be hesitant to dive into a full-blown hierarchical model with several predictors without having a firm handle on how (predicted) means are connected to (noisy) observed data. The suggestion to try a smaller version of your problem wasn’t to “force a fit”, but to allow you to learn more about what your data is like.