what does this code mean?

[River Huang]

Dear All, I found this code (from a paper)

xi: reg dly lly llk lldebts dlk dldebts ///
    i.nwbcode|dlyT i.nwbcode|llyT i.nwbcode|llkT i.nwbcode|lldebtsT i.nwbcode|dlkT i.nwbcode|dldebtsT i.nwbcode, vce(bootstrap)

I wonder if anyone can tell me what does the second line mean? Thanks.

[Scott Merryman]

From the help for xi:

i.varname1|varname3 creates dummies for categorical variable varname1 and continuous variable varname3:
all interactions and main effect of varname3, but no main effect of varname1

Note the difference between the following:

webuse grunfeld
xi: reg invest i.year|kstock
xi: reg invest i.year*kstock

[Clyde Schechter]

Scott is correct. For what it's worth, you would probably be better off abandoning the -xi:- prefix and recasting the regression command in terms of factor variable notation. Which would be:

reg dly lly llk lldebts dlk dldebts i.nwbcode## c.(dlyT llyT llkT lldebtsT dlkT dldebtsT), vce(bootstrap)

(Yes, this code, which uses the ## does generate indicators for nwbcode, but the original equation contains a separate i.nwbcode term, so the distinction between | and * in -xi- that Scott correctly points out is neutralized anyway here.)

[River Huang]

Hi, Scott: Thanks for your explanation.

[River Huang]

Hi, Clyde, Thanks for the suggestion. I will try it out.