Closed bvkrauth closed 1 year ago
Code:
* Setup
sysuse census, clear
* Run the linear regression, with factor and/or time series variables
reg pop marriage i.region
* RCR will not allow this variable list
noisily capture rcr pop marriage i.region
* To fix this, use the Stata command fvrevar to expand the factor/ts variables.
* fvrevar will:
* 1. Create temporary variables for all of the factor levels and lagged variables
* 2. Return the variable list (in r(varlist) with these new temporary variables
* in place of the factor/ts variables
fvrevar marriage i.region
* Copy the new variable list into a macro
local vars = r(varlist)
* See what's in there
di "`vars'"
* Note that fvrevar includes all factor levels including the base level,
* so our variable list is collinear
* We can fix that by using the _rmcoll command, which creates a new
* variable list that omits collinear variables
_rmcoll `vars', forcedrop
* Copy the new variable list into a macro
local vars = r(varlist)
* See what's in there
di "`vars'"
* Now we can use this variable list in the rcr command
rcr pop `vars'
Currently the Stata command rejects varlists that include factor variables or time series operators. In principle, it should be possible to expand those varlists out and use them.