Estimation & Data
- With few repeated observations (only 2 for the outcome) we’ll normally use wide format (see Table 10.1)
Table 10.1: Wide format
treatment
|
outcome_t1
|
outcome_t2
|
covariate1_t1
|
covariate2_t1
|
1
|
15.00
|
21.5
|
1
|
0
|
0
|
8.50
|
10.5
|
0
|
1
|
1
|
15.00
|
27.0
|
1
|
0
|
1
|
24.00
|
23.0
|
0
|
0
|
0
|
13.75
|
11.5
|
0
|
1
|
0
|
40.50
|
24.0
|
1
|
0
|
- Estimation with wide-format data
- \(y_{t} - y_{t-1} = \beta_{0} + \beta _{1} D + \beta X + \varepsilon\)
lm((outcome_t2 - outcome_t1) ~ treatment + covariate1_t1 + covariate2_t1, data = data)
- …we’ll take this approach in the lab below
- Estimation with long-format data