For the answer to this questions, we will use the independent variables mxtsch (total school expenditure), totminc (total monthly income), and foodwage (household food subsidy). Also, we need to clean up the data, keeping only those observations where rel_head==1, and dropping all observations where our independent variables have either missing or invalid (< 0) values.
Once the data is cleaned up, enter the following in the Stata Command window:
reg mxtsch totminc foodwage
This results in the following table:
Source | SS df MS Number of obs = 4196 ---------+------------------------------ F( 2, 4193) = 95.24 Model | 6668804.88 2 3334402.44 Prob > F = 0.0000 Residual | 146804937 4193 35011.9097 R-squared = 0.0435 ---------+------------------------------ Adj R-squared = 0.0430 Total | 153473742 4195 36584.9207 Root MSE = 187.11 ------------------------------------------------------------------------------ mxtsch | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---------+-------------------------------------------------------------------- totminc | .0076642 .0005622 13.632 0.000 .006562 .0087664 foodwage | -.0828563 .041451 -1.999 0.046 -.1641221 -.0015904 _cons | 40.3553 3.249525 12.419 0.000 33.98451 46.72609 ------------------------------------------------------------------------------
Observing the coefficients for the independent variables totminc and foodwage, we can see that with a 1 Rand increase increase in total monthly income, a household will spend .007 Rand more on school expenditures. We can also see that, controlling for total monthly income, with a 1 Rand increase in food subsidy, a household will spend .08 Rand less on school expenditures. (Note that in this example, we have treated the variables as individual-level variables. Was this correct?? No, unless you purposely wanted to give larger households more weight in your answer.)