Nick Cox replied:
> Why? I guess a developer thought it redundant,
> or didn't think it important, or forgot, or something.
> How about
>
> . di sqrt(e(rss)/e(df_r))
>
> or
>
> . di sqrt($S_E_sse/$S_E_tdf)
I was able to replicate this with another (-wagepan-) dataset...
. areg lwage exper married south union d82 d83 d84 d85 d86 d87, absorb(nr)
Number of obs = 4360
F( 10, 3805) = 77.96
Prob > F = 0.0000
R-squared = 0.6160
Adj R-squared = 0.5601
Root MSE = .35324
----------------------------------------------------------------------------
lwage | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-----------+----------------------------------------------------------------
exper | .1130036 .0214828 5.26 0.000 .0708846 .1551227
married | .0577281 .0183605 3.14 0.002 .0217307 .0937254
south | .1107554 .0482741 2.29 0.022 .0161099 .2054009
union | .0843624 .0194332 4.34 0.000 .0462618 .122463
d82 | -.0598641 .0370691 -1.61 0.106 -.1325414 .0128131
d83 | -.1301425 .0566215 -2.30 0.022 -.2411539 -.0191311
d84 | -.1748378 .0771744 -2.27 0.024 -.3261449 -.0235307
d85 | -.2391411 .0981043 -2.44 0.015 -.4314832 -.046799
d86 | -.2937814 .1192146 -2.46 0.014 -.5275121 -.0600506
d87 | -.3463798 .1404224 -2.47 0.014 -.6216903 -.0710694
_cons | .9837128 .0777398 12.65 0.000 .8312972 1.136129
-----------+----------------------------------------------------------------
nr | F(544, 3805) = 8.974 0.000 (545 categories)
. di sqrt(e(rss)/e(df_r))
.35323616
...but not for my own model...
. areg ldmch el1983-el2001 c_difftout c_enp incumb c_normcspd c_normlspd
c_normdspd c_sqldmspd c_cdmargin c_ldmargin c_class ldmseats if !inlist
(_n, 379) [pw=weight], absorb(pano) cluster(region)
Regression with robust standard errors Number of obs = 3347
F( 16, 21) = 478.30
Prob > F = 0.0000
R-squared = 0.7757
Adj R-squared = 0.7276
Root MSE = 3.7027
(standard errors adjusted for clustering on region)
----------------------------------------------------------------------------
| Robust
ldmch | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-----------+----------------------------------------------------------------
el1983 | 11.06352 .7317441 15.12 0.000 9.541772 12.58526
el1987 | 4.070268 .889115 4.58 0.000 2.221252 5.919284
el1992 | 1.856636 1.095025 1.70 0.105 -.4205945 4.133866
el1997 | 3.110942 1.094698 2.84 0.010 .8343927 5.387491
el2001 | 6.117755 1.414653 4.32 0.000 3.175823 9.059687
c_difftout | .0439656 .0249076 1.77 0.092 -.0078325 .0957638
c_enp | 2.263927 .7102512 3.19 0.004 .7868788 3.740975
incumb | -.5054217 .2286173 -2.21 0.038 -.9808573 -.0299861
c_normcspd | -2.735679 .7467913 -3.66 0.001 -4.288716 -1.182641
c_normlspd | -4.763069 .774382 -6.15 0.000 -6.373485 -3.152654
c_normdspd | 1.946096 1.479461 1.32 0.203 -1.130613 5.022804
c_sqldmspd | 5.467829 1.280315 4.27 0.000 2.805268 8.13039
c_cdmargin | .2592689 .0303518 8.54 0.000 .1961488 .322389
c_ldmargin | .1967372 .0247946 7.93 0.000 .145174 .2483003
c_class | -.0103208 .0342645 -0.30 0.766 -.0815777 .0609361
ldmseats | 5.124764 .757683 6.76 0.000 3.549076 6.700451
_cons | -4.274048 .7438351 -5.75 0.000 -5.820938 -2.727158
-----------+----------------------------------------------------------------
pano | absorbed (576 categories)
. di sqrt(e(rss)/e(df_r))
42.40995
The reason is because I switched the -cluster()- option on in my model.
Switching it off and then running that calculation -display-s the RMSE.
I've no idea why the option makes it return a completely different number.
Notice also that the first model produces F statistics on the fixed
effects. Now I've never seen that reported before in -areg-, for it
certainly hasn't appeared in any of my models: it simply says 'absorbed'
every time. Generating dummies for my -pano- variable didn't do the trick
CLIVE NICHOLAS |t: 0(044)7903 397793
Politics |e: [email protected]
Newcastle University |http://www.ncl.ac.uk/geps
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
