Using xi in regress vs. xtreg

[Gerhard Spakowski]

Hi,

I'm running a series of regressions. The basic version is it the form of:
xi: regress y x1 x2 ... x30 i.year i.industry, cluster(ticker)
where ticker is a company's stock ticker (the data has measurements from multiple firms across time).

Now, I also wanted to do a random-effects and fixed-effects version and, based on the help function, I tried the following:
xi: xtreg y x1 x2 ... x30 i.year i.industry, re cluster(ticker)

However, I get the following error:
"i() required" (same error for the fe version)

What am I missing?

[Rich Goldstein]

first, unless you have a quite old version of Stata, do not use "xi" - instead use factor variable notation; see -h fvvarlist-

second, you have apparently not xtset your data (see -h xtset-); Stata needs a way to recognize panels and this can generally be done either by -xtset- used once or for each estimate by using the "i()" option to tell Stata what the panel identifier is

[Gerhard Spakowski]

first, unless you have a quite old version of Stata, do not use "xi" - instead use factor variable notation; see -h fvvarlist-

second, you have apparently not xtset your data (see -h xtset-); Stata needs a way to recognize panels and this can generally be done either by -xtset- used once or for each estimate by using the "i()" option to tell Stata what the panel identifier is

Thanks for the reply.

I run 9.2 and given that the fvvarlist command doesn't show up in the help window, it's not available.

As far as xtset is concerned, when I have multiple firm measurements over multiple years, would "xtset firm year, yearly" be the correct command? How does this affect the usage of the regression functions specified above?