Stata xi

The xi prefix command expands categorical variables into indicator variable sets.

Usage

use http://stata-press.com/data/r14/auto, clear
xi: regress mpg i.rep78

This is equivalent to running something like:

levelsof rep78, local(levels)
foreach l of local levels {
  generate dummy_`l' = rp78 == `l'
}

regress mpg dummy_2 dummy_3 dummy_4 dummy_5

Syntax

The primary use of xi is to take a command like:

xi: regress y i.x1

...and expand the categorical variable x1 into a set of indicator terms for each category. By default, these would be named _Ix1_1, _Ix1_2, and so on. (These variables will persist.) The first such indicator is excluded from the underlying modeling command.

xi also can create interaction terms. For example:

xi: regress y i.x1*i.x2

This creates an indicator for all interactions and all main effects.

If x2 were a continuous variable instead, the more appropriate syntax is one of:

xi: regress y i.x1*x2

...which again creates an indicator for all interactions and all main effects; or...

xi: regress y i.x1|x2

...which again creates an indicator for all interactions and the main effect of x2, but omits the main effect of x1.

See also

Stata manual for xi

CategoryRicottone