Hello,

For the following regression output with the following variables:

• math_scr - average math test score
• comp_stu - number of computers per student
• meal_pct - poverty status (as measured by percentage of students eligible for reduced price school lunches)
• expn_stu - expenditures per student
• avginc - average family income

. regress math_scr comp_stu expn_stu meal_pct avginc, robust

Linear regression Number of obs = 420

F(4, 415) = 281.58

Prob > F = 0.0000

R-squared = 0.7179

Root MSE = 10.005

------------------------------------------------------------------------------

| Robust

math_scr | Coef. Std. Err. t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

comp_stu | 18.74164 9.005385 2.08 0.038 1.03978 36.44349

expn_stu | .0006539 .0009591 0.68 0.496 -.0012314 .0025392

meal_pct | -.4451579 .0249925 -17.81 0.000 -.4942855 -.3960303

avginc | .622685 .1009264 6.17 0.000 .4242943 .8210756

_cons | 657.7018 4.563467 144.12 0.000 648.7314 666.6722

------------------------------------------------------------------------------

A high-ranking Department of Education official publicly advocates the position that academic performance is mainly determined by socioeconomic characteristics and not school resources, and she points to your regression analysis as evidence. She is using your regression results from this question as evidence, citing higher t-statistics (in magnitude) onmeal_pctandavginc.Explain to the Congresswoman what these higher t-statistics do and do not say about the association between socioeconomic variables and math test scores and causation.

My questions are:

1/ What do robust standard errors do?

2/ What do higher t-statistics do / do not say about the association between variables?

Thank you very much!!