Esper: create window for events that occur after another event - complex-event-processing

Question

I want mean speed (sensor=1) at different power settings 1-10 (sensor=2). I don't want speeds from sensor one to be included in the mean calculation if they occurred before the change in setting (sensor=2).
E.g. power_setting=1, speed=50, speed=60, power_setting=2, speed=100, speed=120
If I took a time window around power_setting = 2, then it may include a speed for power_setting = 1.
How do I explicitly tell esper to only calculate with values that occured after the power setting change.
Here is my current script:
CREATE window SpeedWindow.win:time_batch(30 sec) as (speed double, power_setting int);
INSERT into SpeedWindow
SELECT
speedEvent.value as speed,
powerSetting.value as power_setting
FROM
Sensors(id = 1).std:lastevent() as speedEvent,
Sensors(id = 2).std:lastevent() as powerSetting;
INSERT into Output
SELECT
case
when SpeedWindowEvent.power_setting = 1 then 4154
when SpeedWindowEvent.power_setting = 2 then 4155
... etc ...
end as id,
avg(SpeedWindowEvent.speed) as value
FROM
SpeedWindow as SpeedWindowEvent
GROUP BY
SpeedWindowEvent.power_setting;

Answer 1

Based on the comments clarifying the use case. Create a context that initiates and terminates on speed event.
// make it easy to read, takes no CPU at all
insert into Speed select * from Sensor(id=1);
insert into Power select * from Sensor(id=2);
create context SpeedSession initiated by Speed terminated by Speed;
context SpeedSession select avg(...) from Power;