making ri2 play well with high-dimensional fixed effects models

[vbehnam]

I was wondering if there is any way to make ri2 play well with high-dimensional fixed effects packages (primarily lfe, but I'm open to using fixest or plm or any other package). Here is my code and the error message outputted.

own_full_nospill <- lfe::felm(own ~ treat | year:race_fac + race_fac:statefip:county + year:statefip:county| 0 | 0, data=full, weights=full$pop)

N_full <- nrow(full)
m_full <- nrow(full[full$treat == 1, ])
declaration <- declare_ra(N=N_full, m = m_full)

And the output is...

ri2_out_2nofactors <- conduct_ri(formula = own_full_nospill$formula,
                                  declaration = declaration,
                                  sharp_hypothesis = 0,
                                  data = full,
                                  assignment = "treat",
                                  outcome = "own")`

`Error in prep_data(data = data, se_type = se_type, clustered = clustered,  : 
  dims [product 4] do not match the length of object [30496]
In addition: Warning messages:
1: In Ops.factor(year:race_fac, race_fac:statefip:county) :
  ‘+’ not meaningful for factors
2: In Ops.factor(year:race_fac + race_fac:statefip:county, year:statefip:county) :
  ‘+’ not meaningful for factors

[acoppock]

In principle, the formula argument could be passed to any modeling function. this would be a great addition. thanks for the idea.