Hi, I keep getting bad initial energy using the ‘ordered’ transform.
My model is as follows
y0=t_i \ cos(\theta) + \sigma_i
y1=b+ t_i \ sin(\theta) + \sigma_i
I want to add the additional constraint that t is increasing for my datapoints, so I use the ‘ordered’ transform on my t variable. The model looks like this:
For some reason this always gives me a bad initial energy, even when I force my test point to be one where the test value logps are large. None of them are inf, -inf, nan or very low.
Without the ordered transform, the model runs properly.
How would I rectify this?