# st: RE: pseudo r2

 From "Nick Cox" To Subject st: RE: pseudo r2 Date Fri, 7 Nov 2008 18:45:11 -0000

```Stata's definition appears to be 1 -  e(sum_adev) / e(sum_rdev) in this
case.

That's unlike any other definition of pseudo-R^2 that's popular, but it
is natural for a method based on L1-norm (or whatever notation you use).

This underlines one of several problems with these measures: there are
so many to choose from.

FAQ     . . . . . . . . . . . . . . . . . . . . . . . Do-it-yourself
R-squared
9/03    How can I get an R-squared value when a Stata command
does not supply one?
http://www.stata.com/support/faqs/stat/rsquared.html

FAQ     . . . . . . . . . . . . . . . . . . . . . . . . . R-square after
xtgls
4/03    Why does xtgls not report an R-square statistic?
http://www.stata.com/support/faqs/stat/xtgls2.html

FAQ     . . . . . . . . . . . . . . . . . . . .  Missing R-squared for
2SLS/IV
1/03    For two-stage least-squares (2SLS/IV/ivreg) estimates,
why
is the R-squared statistic not printed in some cases?
For two-stage least-squares (2SLS/IV/ivreg) estimates,
why
is the Model Sum of Squares sometimes negative?
For two-stage least-squares (3SLS/IV/ivreg) estimates,
why
are the R-squared and Model Sum of Squares sometimes
negative?
http://www.stata.com/support/faqs/stat/2sls.html

FAQ     . . . . . . . . . . . . . . . . . . . .  Pseudo R-squared for
xtprobit
10/01   How can I calculate the pseudo R-squared for xtprobit?
http://www.stata.com/support/faqs/stat/xtprobit.html

FAQ     . . . . . . . . . . . . . . . . . . . . . . . . . Pseudo-R^2 for
tobit
6/97    Why is the pseudo-R^2 for tobit negative or greater
than 1?
http://www.stata.com/support/faqs/stat/pseudor2.html

FAQ     . . . . . . . . . . . . . . . . . .  R-squared:  areg versus
xtreg, fe
9/96    Why isn't the calculation of R2 the same for areg
and xtreg, fe?
http://www.stata.com/support/faqs/stat/xtr2.html

Nick
n.j.cox@durham.ac.uk

markus.klemusch.m.k@gmx.de

I run a median-regression (qreg) and I wanted to save the "pseudo r2" by
using commands like e() or local
but without success.

gen r2=e(r2) is fine is case of OLS (reg[, robust])
How can I do the same or something similar after running a
median-regression?

People express doubt and concern about the pseudo r2 (according to the
internet).
Is it a second best solution or even a bad solution?

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```