I'm using plm package to analyze unbalanced panel data. The problem is that some firms in the sample have only one year observation. This caused a problem whenever I use the following function (the purpose is to calculate white standard error):
vcovHC(pooling, method="white1", type="sss", cluster="group")
The error information is:
Error in crossprod(X, E(u, ul)) : non-conformable arguments
I'm pretty sure that this is caused by singletons in the data set. But the following call with a different argument causes no problem (the purpose is to calculate cluster robust standard error):
vcovHC(pooling, method="arellano", type="sss", cluster="group")
This works well.
I didn't find any solutions for this problem.
I wonder what's the proper method to analyze unbalanced panel data with singletons or whether I should drop those singletons.
If unbalanced panel data with singletons has no problem, then why does the vcovHC() function in plm package deal with this case with error?