Interaction among variables included in a global

[Ana Moro]

I have a list of variables grouped into two globals

global x ="x1 x2 x3 "
global z="z1 z2"

I need to make the regression of y x1 x2 x3 z1 z2 x1*z1 x1*z2 x2*z2 x2*z1 x3*z1 x3*z2 (so all variables included in globals and their interactiosn)

I prove:

regress y ${x}##${z}

but there is an error and STATA says: "# invalid name"

Any help would be much appreciated. Unlike the example, I have more than five explanatory variables.

Ana

[Clyde Schechter]

sysuse auto, clear

local x mpg headroom trunk
local y weight length

regress price c.(`x')##c.(`y')

You can also do this with global macros if you have some good reason for that choice.

Note the c. prefixes: variables that appear in interactions are, by default, treated as discrete. So on the assumption that your intent is that the variables be continuous, I have added the c. prefixes. If you intend them to be discrete, then just omit the c. prefixes--but be sure to retain the parentheses surrounding the dereferenced macros.

It is OK for the x's to be discrete and the y's continuous or vice versa. However, you cannot do this using a mix of continuous and discrete variables within x or within y.

[Ana Moro]

Thanks a lot. It works