Hi, everyone.
I'm trying to apply Clément de Chaisemartin's DiD model for continuous treatment and different paths with dynamic effects using the did_multiplegt_dyn command. However, I'm encountering some issues. The code that I used is:
did_multiplegt_dyn ent015ua chave_county monthyear auxiliar75, effects(6) placebo(1) effects_equal continuous(1) controls(tent_1) bootstrap(2, 157)
Y: ent015ua (continuous)
ID: chave_county
Time: monthyear
D: auxiliar75 (assumes three values: 1 if the treatment in period t is greater than in the baseline, 0 if equal, and -1 if lower, as pointed out by Chaisemartin GitHub)
tent_1: is a variable that results from the interaction of the treatment baseline with the time variable (monthyear). The baseline treatment is the same for all IDs (0) but is continuous for other periods. Therefore, the variable tent_1 will always be equal to zero in my model.
The problem is this Stata output:
Some control variables are not taken into account for groups with baseline treatment equal to: [0]
This may occur in the following situations:
1. For groups with those values of the baseline treatment, the regression of the outcome first difference on the controls' first differences and time fixed effects has fewer observations than variables.
Note that for each value of the baseline treatment, those regressions are estimated among (g,t)s such that g has not changed treatment yet at t.
2. For groups with those values of the baseline treatment, two or more of your control variables are perfectly collinear in the sample where the regression is run, for instance because those control variables do not vary over time.
This message appears even when I do not use controls. I do not know why. Is this a regular message from this command, or is my code incorrect?
I'm trying to apply Clément de Chaisemartin's DiD model for continuous treatment and different paths with dynamic effects using the did_multiplegt_dyn command. However, I'm encountering some issues. The code that I used is:
did_multiplegt_dyn ent015ua chave_county monthyear auxiliar75, effects(6) placebo(1) effects_equal continuous(1) controls(tent_1) bootstrap(2, 157)
Y: ent015ua (continuous)
ID: chave_county
Time: monthyear
D: auxiliar75 (assumes three values: 1 if the treatment in period t is greater than in the baseline, 0 if equal, and -1 if lower, as pointed out by Chaisemartin GitHub)
tent_1: is a variable that results from the interaction of the treatment baseline with the time variable (monthyear). The baseline treatment is the same for all IDs (0) but is continuous for other periods. Therefore, the variable tent_1 will always be equal to zero in my model.
The problem is this Stata output:
Some control variables are not taken into account for groups with baseline treatment equal to: [0]
This may occur in the following situations:
1. For groups with those values of the baseline treatment, the regression of the outcome first difference on the controls' first differences and time fixed effects has fewer observations than variables.
Note that for each value of the baseline treatment, those regressions are estimated among (g,t)s such that g has not changed treatment yet at t.
2. For groups with those values of the baseline treatment, two or more of your control variables are perfectly collinear in the sample where the regression is run, for instance because those control variables do not vary over time.
This message appears even when I do not use controls. I do not know why. Is this a regular message from this command, or is my code incorrect?
