|
Field
Sample size (or number of clusters) by treatment arms
|
Before
The proportion of spending going to child versus family is randomized, so all units are treated and control at the same time. We could think of the alternative with the lower proportion going to the child support-eligible child as being the control arm, in which case we have 1000 observations in control and 1000 in treatment.
|
After
The proportion of spending (a continuous variable) going to child versus family is drawn from a normal distribution in this experimental design. Taking the alternative with a lower proportion going to the child support-eligible child as being the control arm, we have 1000 observations in control and 1000 in treatment.
|