Hi, I would like to use PYMC3 to sample the relationship between two observed RVs that follow this time series format where there is a time delay between the two variables on a given day:
| | Observations 1 ( t ) | Observations 2 (t + 1) |
|------------|----------------------|------------------------|
| 2021-01-01 | -.1 | -.1 |
| 2021-01-02 | .15 | -.13 |
| 2021-01-03 | .2 | .24 |
I’ve determined that the range for each variable is from between -1.5 and 2 for Observation 1 and -1.2 and 1.8 for observations 2. The shape of the random variables take on something similar to normal distribution with long tails with a mu=0, so I am using the VonMises distribution for both priors.
My goal is to use the observation 1 value to predict a range of values and probabilities for observation 2.
I would like to know how I can set up my priors properly so that .sample() would generate a posterior that would allow me to take observation 1 as an input value and query the posterior such that I can make predictions about observation 2. Do I use a PPC to ensure the accuracy of the posterior for predictions and also to query the posterior in this way? Is there a better way to query the posterior given a value of observation 1?