Need help on this multiple regression calculation

[Aansa Usmani]

Hello,

I was doing this multiple regression calculation for my stats class, and for some reason, I keep having this pop up for educ (education) and inck (income)

reg realinc educ inck

Source | SS df MS Number of obs = 2,146
-------------+---------------------------------- F(2, 2143) = .
Model | 2.0859e+12 2 1.0430e+12 Prob > F = .
Residual | 0 2,143 0 R-squared = 1.0000
-------------+---------------------------------- Adj R-squared = 1.0000
Total | 2.0859e+12 2,145 972455973 Root MSE = 0

------------------------------------------------------------------------------
realinc | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
educ | -4.56e-06 . . . . .
inck | 10000 . . . . .
_cons | -.0000208 . . . . .
------------------------------------------------------------------------------

[Jared Greathouse]

I don't understand.

[Aansa Usmani]

Here’s a picture to get a better understanding of the issue

[Jared Greathouse]

You're measuring the same thing as you're predicting. Hence, you've no need for standard errors or their counterparts

[Aansa Usmani]

is this normal?

[Jared Greathouse]

Well, no. Let me give an example. Say you're a sports team. You win 20 games and lose 20 games one season, and win 15 and lose 15 the next. Your goal is to explain how many games you won. You can't include how many you lost as a predictor because there's a perfect relationship between these. It violates the multicollinearity assumption. In fact, you have PERFECT collinearity in this case. Otherwise, you couldn't have a 100% R squared or a 0 Residual score, this just doesn't ever happen in real life.

The issue here is that you've coded one of these variables improperly

[Aansa Usmani]

How can I tell if there's been a variable that's been coded improperly?

[Aansa Usmani]

How can I tell if there’s been a variable that’s been coded improperly?

[Nick Cox]

Cross-posted and answered at https://www.reddit.com/r/stata/comme...ression_issue/

Please note our cross-posting policy which is that you well us about it. https://www.statalist.org/forums/help#crossposting

Otherwise, there is nothing improper about having two versions of the same variable. What isn't helpful is to use both in a regression.