My objective was to obtain random and fixed models using the PLM function in R. Doing a stationarity check the first difference in all my variables was stationary. So I went ahead and created a model, successfully without dropping any NA values as the plm function by default drops NA values
Just as a double check I created a new panel which contained the first difference of all of my variables and had no NA values.
When Iam trying to estimate a random and fixed effects models using the new panel which has first differenced data and no NA values, Iam getting different results. I want to know what is generally considered the best practice. Do you continue with your original data knowing that plm will and in my case did drop the NA values. Or do you transform the original data and then proceed with modelling. Some variables which in the earlier case (not changing the base panel data) which were not coming significant are now significant (after transforming the base panel data and removing the NA values)
Any help or tips shall be highly appreciated. Iam trying to build a relational panel model, nothing fancy