Parallel dimensions...? Utilizing multiple datetime dimensions for different parameters

Questions v5
,
2

In your initial model, you only have 7 different dow_effect values (I assume one per weekday), so what I would do with that is update the model to:

coords = {
    "weekday": ["Monday", "Thursday", "Wednesday", # continue with rest of days so it is a list of length 7
}
with pm.Model(coords=coords) as model1:
    # priors
    sigma = pm.HalfStudentT("sigma", 4, 1)
    dow_effect = pm.Normal("dow_effect", mu=0, sigma=7.14, dims="weekday")
    # the dims="weekday" will repeat/batch the normal to get a variable with 7 dims,
    # so you can skip the repeat part now
    slope = pm.Normal("slope", mu=0, sigma=.0109)

    # likelihood
    likelihood = pm.Normal(
        "y", 
        mu=(
            slope * ds1["day"].values
            + dow_effect[ds1["weekday"].values]
        ), 
        sigma=sigma, 
        observed=ds1["y"]
    )

It might also be possible to add some dimensions to the likelihood, but I don’t know what these would be.