The issue I’m having is that I would like to pass in an array of counts and then perform the manipulation
x = np.vstack([counts[i:i+W] for i in range(N)])
y = counts[W:]
so that each element of x corresponds to an element of y for linear regression. So ideally I would be able to define a multivariate distribution over the entire array that performs this manipulation to get the x and y data before calculating the likelihood. Does this make sense? Is there a way to define a custom multivariate distribution?