Wald chi2 missing in ivregress without intercept

[Nadiia Lazhevska]

Dear Statalist,

When running ivregress with and without a constant, I noticed that in case of noconst option the Wald chi2 and R-squared are missing.
Why would this happen?

Below is the simplest possible example using a random sample:

. ivregress 2sls Y (X=Z)

Instrumental variables (2SLS) regression          Number of obs   =         20
                                                  Wald chi2(1)    =       3.77
                                                  Prob > chi2     =     0.0520
                                                  R-squared       =     0.1856
                                                  Root MSE        =     .47902

------------------------------------------------------------------------------
           Y |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
           X |  -.1836195   .0945136    -1.94   0.052    -.3688627    .0016238
       _cons |   .2170572   .1666594     1.30   0.193    -.1095892    .5437037
------------------------------------------------------------------------------
Instrumented:  X
Instruments:   Z


. ivregress 2sls Y (X=Z), noconst

Instrumental variables (2SLS) regression          Number of obs   =         20
                                                  Wald chi2(1)    =          .
                                                  Prob > chi2     =          .
                                                  R-squared       =          .
                                                  Root MSE        =     .50321

------------------------------------------------------------------------------
           Y |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
           X |  -.0883567   .0638165    -1.38   0.166    -.2134346    .0367213
------------------------------------------------------------------------------
Instrumented:  X
Instruments:   Z

.

Y    X    Z
.4585037    .5176533    .3329265
.7285853    1.191534    .9394104
.9345604    1.533666    1.430296
.1397186    1.119215    .9374179
.0714826    .4621876    .3363032
.2510669    .3417048    .2687683
-1.014193    1.273385    1.263353
-.5596751    .6984423    .3172847
.0162329    .3824131    .3631766
.113968    .6447593    .6223881
.3675321    .4940272    .4221866
.0183036    .6342704    .5214227
.2845293    1.301095    1.174958
-.0480449    4.518545    3.468179
.150546    1.301085    1.495178
-.2118224    1.452726    .9303899
-.4447729    2.156844    1.523062
-1.33329    4.518545    3.468179
.034584    1.385382    1.09914
-.5777966    1.091038    .8542542

Please note that this question is cross-referenced on StackExchange:
https://stats.stackexchange.com/ques...vregress-stata

[Phil Bromiley]

I think lots of folks do not see R-square as that meaningful without an intercept. There are ways to calculate it - see https://stats.stackexchange.com/ques...2-in-linear-mo
If you really want an R-square, you can always do the predicted values and use the formula in this stackexchange discussion.

[Nick Cox]

Echoing Phil, http://www.stata.com/support/faqs/statistics/r-squared/ spells this out in excruciatingly elementary terms.

[Nadiia Lazhevska]

Phil and Nick, thank you for very useful links