coords = {
"date": dates,
"channel": channels,
"state": mn_state,
"cohort": mn_cohort,
}
where
dates = [ 0 1 2 ... 99]
channels = ['TV_val']
mn_state = Index(['Queensland', 'Victoria', 'WA', 'NT'], dtype='object')
mn_cohort = Index(['Males', 'Females', 'Other'], dtype='object')
If it makes any difference, the more complex model is using an adstock and saturation function on the channel data from the pymc_marketing package.