ToDo.md

Things to Do

Better Make_*_Data Options

add focal error_type for make_spacetime_data

• include option @param sar_factor, ar_factor parameter for error_type='focal'

add better weights_matrix functionality for make_time_data

• also, use stats::arima.sim, stats::filter

make_space_data has two types of "focal" errors: SAR, SMA

Better vcov* Options

• different kernels (currently only Bartlett)
• small sample adjustments
• add cluster correction via sandwhich::cluster
• option to calculate space weights-matrix
• option to calculate time weights-matrix
• option to calculate space-time kronecker weights-matrix

Better Support for non-lm objects

lfe::felm

• SHACsep with heteroskedasticityy FAILS for felm objects with type='HC3'
• SHACsepHC=vcovSHACsep(reg, method='bruteforce_xy',cutoff_s=.6, loc_x=DFst$Loc_X, loc_y=DFst$Loc_X, loc_t=DFst$Time_ID)

plm

Support Bootstrapping

This is where pre-computing and recycling a weights matrix is most useful

More Documentation (README or Vignette)

provide real example (e.g., stata example)

• http://freigeist.devmag.net/economics/936-spatial-temporal-autocorrelation-correction-with-r-conley.html
• https://github.com/darinchristensen/conley-se

example of bootstrapping efficiency with method='rolled'

Recent Changes

Eliminated data.tables dependancy

Eliminated pre-formatting data-matrices

Name Changes

• vcovSTsep -> vcovST -> vcovSHACsep(add_hc=F, add_hac=T)
• vcovSCL -> vcovST.loop -> vcovSHACsep(add_hc=F, add_cl=T)