Why can't create aggregated event for some aggregations
-> created from the frequency analysis (of the training dataset)
-> when in a PQL model, cannot create aggregated events
aggregated_event_value is a boolean actually : 1 -> has at least x activities
prob conversion if not this
prob conversion if this
phi
n_11 * n_00 - n_10 * n_01) / sqrt(n_1X * n_0X * n_X0 * n_
How to interpret negative phi ?
Nb people done
percent of people who did action
number of occurences of event last x days
If aggregted event= at least 10 events
if add aggregation in the signals, -> are the people where aggregated_event_value = 0 appear
COUNT(DISTINCT CASE WHEN aggregated_event_value = 1 AND target = 1 THEN email ELSE NULL END) AS n_11, | |
| COUNT(DISTINCT CASE WHEN aggregated_event_value = 1 AND target = 0 THEN email ELSE NULL END) AS n_10, |
| COUNT(DISTINCT CASE WHEN aggregated_event_value = 0 AND target = 1 THEN email ELSE NULL END) AS n_01, |
| COUNT(DISTINCT CASE WHEN aggregated_event_value = 0 AND target = 0 THEN email ELSE NULL END) AS n_00, |
| COUNT(DISTINCT CASE WHEN aggregated_event_value = 1 THEN email ELSE NULL END) AS n_1X, |
| COUNT(DISTINCT CASE WHEN aggregated_event_value = 0 THEN email ELSE NULL END) AS n_0X, |
| COUNT(DISTINCT CASE WHEN target = 1 THEN email ELSE NULL END) AS n_X1, |
| COUNT(DISTINCT CASE WHEN target = 0 THEN email ELSE NULL END) AS n_X0, |
| COUNT(DISTINCT email) AS n, |