l*****n 发帖数: 76 | 1 regression 是这样的
R2=a+b*R1+c*V*R1+error
如何show the economic importance of the coefficient c on the interaction
term?
mean of V=0; mean of R1 =0
可以比较the difference in the values of R2 between the case where V is one
standard deviation above its mean and R1 is one standard deviation above its
mean and the benchmark case where V=0 and R1 is one standard deviation
above its mean吗 |
t*****i 发帖数: 68 | 2 No, because in this way the outcome you get includes the direct effect of R1
as well as that of the interaction term. You have to use a diffrence-in-dif
frence approach to show the effect solely from the interaction term.
its
【在 l*****n 的大作中提到】 : regression 是这样的 : R2=a+b*R1+c*V*R1+error : 如何show the economic importance of the coefficient c on the interaction : term? : mean of V=0; mean of R1 =0 : 可以比较the difference in the values of R2 between the case where V is one : standard deviation above its mean and R1 is one standard deviation above its : mean and the benchmark case where V=0 and R1 is one standard deviation : above its mean吗
|
l*****n 发帖数: 76 | 3 why? is the direct effect of R1 removed already by taking the difference
between two cases?
suppose one standard deviation of R1=1
one standard deviation of V=1
case1 R2=a+b*1+c*1*1
benchmark case R2=a+b*1+c*1*0
the difference between case 1 and benchmark case will be Delta_R2=c*1*1
This is the effect of interaction term, right?
R1
dif
【在 t*****i 的大作中提到】 : No, because in this way the outcome you get includes the direct effect of R1 : as well as that of the interaction term. You have to use a diffrence-in-dif : frence approach to show the effect solely from the interaction term. : : its
|
t*****i 发帖数: 68 | 4 My bad. I misunderstood you. Then it's fine to show the effect of the intera
ction this way. However, people ususally also incluce V separately in the re
gression unless there are theoretical reasons for not doing so. you may want
to justify your specification more carefully.
【在 l*****n 的大作中提到】 : why? is the direct effect of R1 removed already by taking the difference : between two cases? : suppose one standard deviation of R1=1 : one standard deviation of V=1 : case1 R2=a+b*1+c*1*1 : benchmark case R2=a+b*1+c*1*0 : the difference between case 1 and benchmark case will be Delta_R2=c*1*1 : This is the effect of interaction term, right? : : R1
|
j**********n 发帖数: 390 | 5 It is more commonly seen as:
R2 = a + b*R1 + c*V*R1 + d*V + error |