Thank you for your help! Would it be possible to see the code you used to produce the RMSE and NLPD plots as I think this would be very useful for my project. Sorry very new to python! And yes I can see what you mean by including the error of the data, for mine it produces a very similar Gaussian distribution plot it just takes the data training points as less certain (as I was having issues with the sampling taking each point as have a zero error) although since introducing trace instead of MAP it seems to be okay so I could most likely run it without that
Also your Gumbi package looks very interesting, so would this perform a similar process to what I have written above but automate it more? And what are the results of the package, how do they cross validate?