Difference in Number of Obs and Regression Model

[Hoa Phan]

Hi guys,
I have the question is that the number of Obs (A variable) is 1,746.
However, when I run regression, tthe number of obs is 1,751.
I do not know the reason why?
Could you please let me know?
Thank you in advance!


sum A

Variable | Obs Mean Std. Dev. Min Max
-------------+---------------------------------------------------------
A | 1,746 2.336577 .6250715 -1.479457 4.588375


r regression Number of obs = 1,751
F(219, 1750) = 5.72
Prob > F = 0.0000
R-squared = 0.2108
Root MSE = .15951

(Std. Err. adjusted for 1,751 clusters in double_cluster1)
--------------------------------------------------------------------------------
| Robust
SemiexcessR | Coef. Std. Err. t P>|t| [95% Conf. Interval]
---------------+----------------------------------------------------------------
A | -.0496081 .0248835 -1.99 0.046 -.0984127 -.0008035

[Carlo Lazzaro]

Hoa:
if you did not impose an -if- qualifier when -sum-, this finding is a mistery.
That said, could you please share what you typed annd what Stata gave you back (the whole code and output). Thanks.

[Hoa Phan]

Dear Carlo,
Yeah, I found my mistakes because I dropped my missing values in a wrong way.
However, I fixed it.
Thank for your valued comment!
Hoa,

[Joro Kolev]

Just so that future visitors of this thread are not mistified:

It is totally normal that when you summarise one variable, say Y, you get more observations than if you do

reg Y X Z W.

For the first the count is the non-missing Ys, and for the second the count is the observations for which all of Y, X, Z, W are not missing. Obviously the second could be only as large as the first, and often it will be smaller if any of X, Z , W is missing for some observations.

Getting results the other way round as OP did is a mistery, and suggest some error in code.

[Hoa Phan]

Yes, it is a good point. Thank Joro!