在stata manual给出的一个案例中:
Assume that we were to collect data by randomly sampling 10,000 doctors (from 100 hospitals)
and then sampling 10 patients of each doctor, yielding a total dataset of 100,000 patients in a cluster
sample. If in some regression we wished to include effects of the hospitals to which the doctors
belonged, we would want to include a dummy variable for each hospital, adding 100 variables to our
model. areg could fit this model by
areg depvar patient vars, absorb(hospital) vce(cluster doctor)
其中,, absorb(hospital) vce(cluster doctor)的分工体现在,absorb控制了医院的固定效应,vce控制了医生层面抽样的相关性。在式中,其为何不控制医生固定效应?
我目前练习的一个例子中,数据unit为跨国层面的企业数据,我想考察国家层面的某一变量对企业创新的影响,我可不可以只控制国家固定效应,不控制产业固定效应,仅用vce(cluster industry)来控制产业内的企业特征相关性?
另外,我们所说的产业内企业的相关性,是针对于我们所说的y,还是x,还是x,y的组合?
谢谢。