A rule of thumb for outliers is to look for values that are:

a) > Q3 + 1.5*(Q3-Q1), or

b) < Q1 - 1.5*(Q3-Q1),

where Q1 is the cutoff for quartiles 1 and 2 and Q3 is the cuttoff for quartiles 3 and 4.

You can do this in Stata as follows:

*begin stata code

qui sum var, d // replace var with your variable

gen outlier = !inrange(var, `=r(p25) - 1.5*(r(p75)-r(p25))', `=r(p75) + 1.5*(r(p75)-r(p25))')

tab outlier, m

*end stata code

Note that this is a rule of thumb only!

Also rember that in most cases you will want to be careful with outliers. Single ones do not violate distributional assumptions severely, and if there’s no very good reason to remove an observation (methodological flaw, error, nonstandard observation) It should not be done lightly.

You can obtain influence statistics such as leverage, Cook's distance, dfbetas using the predict command in STATA after running the regression model and evaluate your results appropriately.