Dear all,
I am trying to measure how the impact of an exogenous indicator variable varies across three different categories, where the assignment of observations into these categories may be endogenous. I consider using CMP (Roodman, 2011) with a code below:
, where I1 =1 if the observation falls into the second category, 0 otherwise. I2=1 if the observation falls into the third category, 0 othewise.
A is the exogenous indicator variable of interest.
I1_A = I1*A, and I2_A = I2*A
I is a categorical variable with values of {0, 1, 2}.
Z1 and Z2 are instruments. Z1sq= Z1*Z1, Z2sq = Z2*Z2, and Z1Z2 = Z1*Z2.
I am wondering whether the model above is valid. Or, do I need to add extra equations with the interaction terms as dependent variables? In other words, do I need to use a code like below?
Thank you for your advice in advance.
Best regards,
Hong
I am trying to measure how the impact of an exogenous indicator variable varies across three different categories, where the assignment of observations into these categories may be endogenous. I consider using CMP (Roodman, 2011) with a code below:
Code:
cmp ( Y = I1 I2 A I1_A I2_A $x i.cbsa) (I = Z1 Z2 Z1sq Z2sq Z1Z2 $x, iia), nolrtest ind($cmp_cont $cmp_mprobit) tech(dfp nr) nonrtol vce(robust) ghkdraws(7) difficult
A is the exogenous indicator variable of interest.
I1_A = I1*A, and I2_A = I2*A
I is a categorical variable with values of {0, 1, 2}.
Z1 and Z2 are instruments. Z1sq= Z1*Z1, Z2sq = Z2*Z2, and Z1Z2 = Z1*Z2.
I am wondering whether the model above is valid. Or, do I need to add extra equations with the interaction terms as dependent variables? In other words, do I need to use a code like below?
Code:
cmp ( Y = I1 I2 A I1_A I2_A $x i.cbsa) (I1_A = Z1 Z2 Z1sq Z2sq Z1Z2 $x) (I2_A = Z1 Z2 Z1sq Z2sq Z1Z2 $x) (I = Z1 Z2 Z1sq Z2sq Z1Z2 $x, iia), nolrtest ind($cmp_cont $cmp_probit $cmp_probit $cmp_mprobit) tech(dfp nr) nonrtol vce(robust) ghkdraws(7) difficult
Best regards,
Hong